Switch to DuckDuckGo Search
   February 1, 2013  
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | >

[00:00:46] *** iyp has quit IRC
[00:01:41] *** estibi_remote has quit IRC
[00:02:50] *** jesser has quit IRC
[00:03:02] *** jesser has joined #smartos
[00:06:14] *** wolfeidau has quit IRC
[00:06:28] *** wolfeidau has joined #smartos
[00:10:09] *** estibi_remote has joined #smartos
[00:21:03] *** potatosalad has quit IRC
[00:22:26] *** Vod has joined #smartos
[00:24:50] *** wolstena has quit IRC
[00:26:00] *** ira has quit IRC
[00:29:30] *** enmand has quit IRC
[00:29:30] *** Cpt-Oblivious has quit IRC
[00:32:43] *** ipalreadytaken has quit IRC
[00:36:29] *** Cpt-Oblivious has joined #smartos
[00:50:39] *** enmand has joined #smartos
[00:50:49] *** szaydel has joined #smartos
[00:57:33] *** leecallen35 has quit IRC
[00:57:50] *** leecallen has joined #smartos
[01:00:24] *** potatosalad has joined #smartos
[01:04:08] *** enmand has quit IRC
[01:10:27] <jesse_> ok, this may be the single malt working it's magic, because I'm just all-wtf
[01:10:34] *** wolstena has joined #smartos
[01:17:59] <jesse_> http://pastebin.com/SESDwtXb
[01:19:01] <rmustacc> jesse_: What's the question based on that data?
[01:19:10] <jesse_> dammit
[01:19:22] <jesse_> still pasted the set*() in wrong order
[01:19:43] <jesse_> though, the setuid value should be 22 then, not 2?
[01:21:20] <jesse_> because the code after that recv() is
[01:21:21] <jesse_>         if(ret < 0 && errno == EAGAIN)
[01:21:36] <jesse_> and it never gets in that if
[01:21:50] <rmustacc> What's the source code in question look like?
[01:22:47] <jesse_> whole function, from xlibxcb: http://pastebin.com/Ms4ueDm5
[01:23:36] <jesse_> switched setuid and setgid, still 2 and -1
[01:24:26] <rmustacc> Those setuid and setgid calls look totally busted.
[01:24:31] <rmustacc> Why would I ever setuid to errno?!
[01:24:52] <rmustacc> In fact the getgid() there also looks wrong.
[01:25:01] <rmustacc> The differeing indentation also leads me to be suspicious.
[01:25:02] *** iyp has joined #smartos
[01:25:09] <jesse_> oh, that's my hack-it-up print to truss output=)
[01:25:47] <jesse_> couldn't remember any other way of getting the values to the right place in the output
[01:26:04] <jesse_> the problem is the values
[01:26:17] <jesse_> why is errno 2? shouldn't it be 11?
[01:26:53] <jesse_> truss seems to think recv returns Err#11
[01:28:31] <rmustacc> Truss at least seems to have reordered the calls.
[01:28:41] <rmustacc> You're doing a setgid(-1);
[01:28:53] <rmustacc> Then a setuid of errno which has changed because of the setgid call.
[01:29:00] <rmustacc> Why they got reordered is a mystery to me though.
[01:29:12] <jesse_> oh, I did that, because cut'n'paste
[01:29:20] <jesse_> I reordered them, and then pasted the code
[01:29:32] <jesse_> also; running again:
[01:29:33] <jesse_> 44976/31:       recv(6, 0x0868DC58, 8, 0)                       Err#11 EAGAIN
[01:29:33] <jesse_> 44976/31:       setuid(2)                                       Err#1 EPERM [proc_setid]
[01:29:36] <jesse_> 44976/31:       setgid(4294967295)                              Err#22 EINVAL
[01:30:11] <jesse_> (because, obviously, errno will be wrong after the first set* call)
[01:31:54] <jesse_> also, in the first case, errno should have been 22? (and hence, the argument to setuid() instead of 2?)
[01:32:18] *** iyp has quit IRC
[01:34:07] <rmustacc> I'm not sure. I would consider using DTrace to verify what's going on.
[01:35:05] <jesse_> is it possible that the program linking with libxcb.so has defined errno?
[01:35:17] <jesse_> and that would cause it not to change?
[01:36:06] <rmustacc> There could be something crazy linking wise going on here, but that wouldn't be my first guess.
[01:36:44] <jesse_> well, the library works with xlogo, and not with java. so link problems are very much on the table=)
[01:36:44] <wesolows> recall that error should expand to a call to ___errno() and a dereference of the return value
[01:36:49] <wesolows> *errno
[01:36:57] <wesolows> you can check the assembly to verify that.
[01:37:05] <wesolows> (this implements the per-thread errno)
[01:37:22] <rmustacc> Right, thanks wesolows.
[01:40:39] *** iyp has joined #smartos
[01:40:40] <jesse_> so, basically, I should only find undef ___errno in nm of the libs?
[01:41:00] <wesolows> undef'ing that symbol would be a severe programming error
[01:41:12] <jesse_> and global errno would be bad
[01:41:27] <wesolows> in a program with multiple threads, it would yield incorrect results
[01:41:33] <wesolows> *process
[01:41:35] *** iyp has quit IRC
[01:41:56] <wesolows> are you certain that recv is returning -1 here?
[01:42:12] <wesolows> after all, you're not setting errno to 0 beforehand, so if recv succeeds you cannot say anything about errno
[01:42:20] <wesolows> dtrace is a great friend here
[01:42:29] <jesse_> yes, 4294967295-2^32 is -1
[01:43:00] *** iyp has joined #smartos
[01:43:00] <jesse_> wesolows, looking at libxcb code, affects every X app=)
[01:43:13] <wesolows> can you paste read_block::dis?
[01:43:16] <wesolows> from mdb
[01:43:52] <wesolows> preferably without your additions
[01:43:56] <jesse_> how would I get that from java app that launches separate java processes
[01:44:08] <wesolows> mdb -p <pid>
[01:44:10] <jesse_> or can I just load the .so somehow in mdb?
[01:44:14] <wesolows> or that
[01:44:27] <wesolows> though in that case it won't have the relocations I want
[01:46:38] *** ipalreadytaken has joined #smartos
[01:49:19] <jesse_> it runs too fast, how do I catch the process?
[01:50:52] <wesolows> dtrace -w -n 'pid$target::main:entry { stop(); }'
[01:51:01] <wesolows> with -p <pid>
[01:51:14] <jesse_> yes, but I don't know the pid
[01:51:14] <wesolows> I don't know how these processes are being created, who's doing it etc.
[01:51:14] *** iyp has quit IRC
[01:51:24] <jesse_> as I run a process that starts another process
[01:51:30] <wesolows> syscall::fork:return ?
[01:51:38] <wesolows> or the proc provider
[01:52:46] <wesolows> it looks like even just attaching mdb to the library may work
[01:52:56] <wesolows> if it's really that hard to attach to the running process
[01:53:16] <jesse_> my dtrace skills are close to 0
[01:54:20] <jesse_> java testing framework is a java process that starts several other processes and after that tells them to run some java code, and restarts them if it thinks it would be prudent before running the next test
[01:54:51] <jesse_> feel free to come up with a simple way of figuring that out=)
[01:54:53] <wesolows> ok.  so let's just feed mdb the library or executable that contains read_block
[01:54:58] <wesolows> I'll try to make do.
[01:56:03] *** iyp has joined #smartos
[01:57:33] <jesse_> can't find itm it's static
[01:57:45] <jesse_> to make things easy=)
[01:57:56] <wesolows> that shouldn't matter.  static symbols should have local visibility.
[01:58:04] <wesolows> in nm parlance, "FUNC LOCL"
[01:58:25] <jesse_> hmmm, true
[01:58:31] <jesse_> lookin wrong library, apparently
[01:58:41] <wesolows> if the build process stripped the binary (which is a bug; this should never be done), you can use nm -D
[01:59:10] <jesse_> libxcb compiles into 22 separate libraries, as a bonus
[01:59:18] <wesolows> although that may only work for globals, hmm
[02:01:55] <jesse_> only libxcb.a has read_block, would that suffice?
[02:02:05] <wesolows> wait, these are static libraries?!
[02:02:14] <jesse_> there are also static libraries
[02:02:20] <jesse_> courtesy of pkgsrc
[02:02:22] <wesolows> well, let's try that
[02:02:29] <wesolows> I'm not sure mdb can read those
[02:03:15] <jesse_> uh, sorry, I lied
[02:03:20] <jesse_> it's not there either
[02:03:38] <jesse_> just misread _xcb_in_read_block which calls read_block
[02:04:18] <wesolows> well, how about using progenyof() to alleviate the unknown pid issue, then trace on pid$target::___errno:entry with a predicate on caller?
[02:04:30] *** potatosalad has quit IRC
[02:04:59] <wesolows> or maybe I should stop trying to be helpful; it's probably just frustrating.
[02:05:06] <jesse_> as I'm re-compiling the library constantly already, if I remove the static from it would it change the linking?
[02:05:23] <wesolows> hopefully it will prevent it from being stripped, yes
[02:05:40] <wesolows> my personal preference is to make /path/to/strip be a link to /bin/true
[02:06:11] <wesolows> there is basically never a situation in which I want a binary to be stripped
[02:06:36] <jesse_> wesolows, http://pastebin.com/ZZM5EaB9
[02:06:42] <jesse_> without my additions
[02:07:20] <jesse_> sorry, one getgid remained=)
[02:07:42] <jesse_> (used earlier to see if that was even the right function)
[02:07:51] <wesolows> winner
[02:07:52] <wesolows> read_block+0x59:                movl   0x74(%ebx),%edx
[02:07:52] <wesolows> read_block+0x5f:                cmpl   $0xb,(%edx)
[02:08:00] <wesolows> your library is being mis-built
[02:08:21] <wesolows> if you want to use it in a multi-threaded process, you must build it with -D_TS_ERRNO or -D_REENTRANT
[02:08:24] <jesse_> wrong -Ds?
[02:08:27] <wesolows> yes
[02:08:34] <jesse_> let's see...
[02:09:10] <wesolows> there is also some combination of -D_XOPEN_SOURCE blah blah blah that can make the right thing happen
[02:09:17] *** doki_pen has left #smartos
[02:09:28] <wesolows> see /usr/include/errno.h; it's quite simple
[02:09:48] <wesolows> the _POSIX_C_SOURCE stuff is in sys/feature_tests.h
[02:09:49] <jesse_> I had to play with -D_XOPEN_SOURCE etc. to get the sun crypto to compile
[02:09:52] <jesse_> picky code
[02:10:25] <jesse_> there was exactly one combination, with __EXTENSIONS__ that would get everything right
[02:10:58] <wesolows> good old __EXTENSIONS__.
[02:11:33] <wesolows> well, -D_TS_ERRNO doesn't interact with anything else, unlike -D_REENTRANT.  I don't recommend using it, but if -D_REENTRANT just won't work for some reason, you could try it to see if this is the problem.
[02:12:00] *** iyp has quit IRC
[02:13:45] <jesse_> https://github.com/joyent/pkgsrc/blob/trunk/x11/libxcb/Makefile
[02:13:54] <jesse_> too many XXXs to my liking
[02:13:58] *** iyp has joined #smartos
[02:15:02] <jesse_> why would be passing -D_REENTRANT be a bad thing?
[02:15:11] <jesse_> it should work with singlethreaded programs the same?
[02:16:16] <wesolows> it should still work, yes
[02:16:26] <wesolows> it's just a bit more expensive.  you absolutely don't care.
[02:16:48] <jesse_> well, you do when it suddenly doesn't work in a multithreaded program
[02:17:17] <wesolows> it does look like it's trying to always pass -D_REENTRANT
[02:19:19] <jesse_> no idea what it's passing, it only prints out 'CC' for compile command
[02:19:25] <wesolows> there's also the possibility that someone is doing #undef errno and then #defining it to something else.
[02:19:32] *** robinbowes has quit IRC
[02:19:33] <wesolows> fucking quiet build
[02:19:39] *** robinbowes has joined #smartos
[02:19:40] <wesolows> you can usually pass make V=1 to fix that
[02:20:25] <jesse_> that worked
[02:21:47] <jesse_> nice, it's calling libtool
[02:22:00] <jesse_> and no -D_REENTRANT on the command line
[02:22:27] <wesolows> any -D_XOPEN_SOURCE?
[02:22:42] <wesolows> if none of those is making it through, then you've found the problem.
[02:22:51] <wesolows> ok.  so this is a pkgsrc bug then, most likely.
[02:23:06] <jesse_> thanks for the help
[02:23:08] <wesolows> or a libtool/xcb bug that's filtering out the necessary defines.
[02:23:16] <wesolows> sure, sorry it took so long
[02:23:29] <jesse_> well, I was staring at the problem for most of the day
[02:23:30] <wesolows> I suspected this from the start, but we need to be sure.
[02:23:46] <jesse_> as it didn't make sense xlogo works and java-invoked code doesn't
[02:24:30] <jesse_> I hate it I always forget dtrace completely between the times I need it
[02:24:42] <jesse_> annoying to re-re-read about it
[02:25:06] <wesolows> fwiw I normally have the guide open
[02:25:26] <wesolows> the syntax is second nature and I know most of the capabilities, but there are a lot of providers and a few subroutines I rarely use.
[02:26:22] <jesse_> still have the java providers to make compile
[02:27:08] <jesse_> actually, they might just work, as I fixed some problems elsewhere in the build that might have affected it
[02:27:16] *** nefilim has joined #smartos
[02:32:43] *** ipalreadytaken has quit IRC
[02:35:32] *** iyp has quit IRC
[02:36:38] <jesse_> CFLAGS didn't work
[02:36:42] <jesse_> it needed CPPFLAGS
[02:37:03] <jesse_> (compiling .c with gcc, so obvious choice=))
[02:37:26] <jesse_> or maybe I did something wrong and the rebuild didn't go the right way the first time
[02:38:09] *** ira has joined #smartos
[02:38:46] <jesse_> and it works
[02:38:54] <wesolows> CPPFLAGS is normally the right answer
[02:39:06] <wesolows> since sometimes things are run through the preprocessor separately
[02:39:15] <wesolows> but I'm glad to hear that it works!
[02:39:21] <jesse_> though, the X problem was hiding some other problem... just like I suspected when I wouldn't give up on X failures=)
[02:39:35] <wesolows> CPP = C preprocessor; C++ is referred to as CXX in make macros
[02:40:13] <rmustacc> I wish everyone just followed CPPFLAGS, CFLAGS, LDFLAGS.
[02:40:20] <wesolows> don't forget LIBS
[02:40:21] <rmustacc> But no such luck.
[02:40:24] <rmustacc> And LIBS, my mistake.
[02:40:35] <wesolows> most things do, but people who try to be too clever invariably fuck it up
[02:40:39] <wesolows> see node.js
[02:40:58] <wesolows> which is #1 by a landslide in angry makefile comments in illumos-extra
[02:41:00] <jesse_> "people who try to be too clever" made the jdk build system
[02:41:12] *** tonyarkles has quit IRC
[02:41:15] <jesse_> I'm afraid to see the configure&&make abomination they've cooked up=)
[02:41:28] <rmustacc> At least it's make .
[02:41:30] <wesolows> whatever it is, I'm sure it won't work.
[02:41:31] <jesse_> (actually, it might be ok)
[02:41:49] <wesolows> right, it's still better than waf-scons-rake-gyp-ant-etc
[02:42:18] <wesolows> those things can take their retarded older brother libtool and go on a nice long vacation in a volcano
[02:42:32] <jesse_> ant is actually pretty good
[02:42:38] <jesse_> as you can't do much with it
[02:42:44] <wesolows> relative to gyp, sure
[02:42:54] <wesolows> and since it uses XML no one is silly enough to try to use it
[02:42:55] <jesse_> then again, there are heroes who manage to extend it
[02:42:58] <wesolows> but still...
[02:43:06] <jesse_> to the point that +-0.0.1 in version breaks it
[02:43:08] <wesolows> Just. Use. Make.
[02:43:30] <jesse_> for java, ant works better
[02:43:45] <jesse_> though I hate it with a passion
[02:43:58] <jesse_> but isn't that the case with C[++] and make
[02:44:02] <wesolows> well, and that's why I'm so dumb it hurts.
[02:44:21] <wesolows> my take is that if the "best" way to build something in a language is to use ant, that oughtta be a sign.
[02:44:40] <jesse_> if you want to step into the world of no return, you touch maven
[02:45:15] <jesse_> it sucks you in, chews you good and then spits you out and you wonder if it would have been easier just to write your own build system instead=)
[02:47:26] <jesse_> ok, I still have one more X related bug, with libXau
[02:47:35] <jesse_> I wonder if I should just defer to jperkin=)
[02:49:03] *** iyp has joined #smartos
[02:50:41] *** iyp has quit IRC
[02:50:59] <jesse_> it does -D_REENTRANT, so not the same easy fix=)
[02:54:34] <jesse_> ah, it's not libXau, it's the test framework stripping the env of $XAUTHORITY
[02:54:47] <jesse_> WONTFIX
[03:05:46] *** Vod has quit IRC
[03:19:41] *** ipalreadytaken has joined #smartos
[03:20:46] *** szaydel has quit IRC
[03:25:54] *** ipalreadytaken has joined #smartos
[03:32:36] *** ipalreadytaken has quit IRC
[03:34:19] *** miine_ has joined #smartos
[03:34:33] *** ipalreadytaken has joined #smartos
[03:34:45] *** darjeeling has quit IRC
[03:36:31] *** dap has quit IRC
[03:37:14] *** miine___ has quit IRC
[03:37:15] *** jamesd has joined #smartos
[03:43:29] *** szaydel has joined #smartos
[03:53:41] *** iyp has joined #smartos
[03:53:56] *** ipalreadytaken has quit IRC
[04:07:20] *** dap has joined #smartos
[04:14:51] *** szaydel has quit IRC
[04:26:40] *** darjeeling has joined #smartos
[04:30:34] *** jwk404 has quit IRC
[04:30:42] *** jwk404 has joined #smartos
[04:32:27] *** Qten has joined #smartos
[04:38:46] <Qten> hey guys, just had this idea, i know probably crazy however when using zfs say for example you have 4 Servers with 6 1TB Disks each, you export the disks using ISCSI into a head node running ZFS, and mirror and stripe the disks across all the available disks would give you reasonable performance i imagine, now lets say you add another 4 servers with 6 1TB disks and do the same thing add these
[04:38:46] <Qten> servers to the zpool in mirrors now we'll have 50% of the total disks for example at 50% Capacity and the new disks at 0% capacity. however we wont be able to use all the iops of all of the disks in the pool. So now to my idea....
[04:39:36] *** artimus has joined #smartos
[04:40:35] <Qten> I parition each of the disks into say 10gb paritions and mirror these across the disks instead and when the new servers get added i add a slice of the new disk to the existing mirror pool turning the 2 way mirror into a 3 way one and once synced i remove one of the old slices there by distrubuting the data across the new disks and using the old slice with a new slice to create a new mirror
[04:40:35] <Qten> set. hence how we would obtain the full iops avaliable over all disks.
[04:41:28] <e^ipi> iscsi will never give you reasonable performance
[04:41:43] <Qten> i was thinking maybe iscsi srp
[04:41:46] <e^ipi> nope
[04:59:07] *** echelog-1 has joined #smartos
[05:00:51] <Qten> example as to what i was talking about http://pastebin.com/v6MkaQtT
[05:00:51] *** marsell has quit IRC
[05:01:48] <wesolows> Qten: seriously, it's not a good idea.  Risk is off the charts, performance will suck, and you're not getting much benefit at all.  It's basically EBS, but worse.
[05:02:00] <Qten> hmm
[05:02:25] <ipAlreadyTaken> Yep just checked and I def have some slow ones
[05:02:28] <wesolows> The use of local storage to back instances is a gigantic win.
[05:02:30] <Qten> how so? zfs should be keeping the data intact and doing the replication
[05:02:49] <wesolows> ZFS can't save you from devices that just "get slow"
[05:03:22] <wesolows> It also can't save you when something inside the iSCSI stack loses its mind.
[05:03:34] <Qten> ok lets say aoe then?
[05:03:35] <wesolows> And it certainly can't alleviate the mind-boggling complexity of such a system.
[05:03:46] <wesolows> are you trolling me?
[05:03:51] <richlowe> weselows: there's been intermittent talk of causing it to try to.
[05:03:52] <Qten> lol no
[05:04:03] <wesolows> you've gone from bad to worse
[05:04:06] <Qten> such systems as coraid use aoe
[05:04:09] <wesolows> ata over ethernet is a fucking joke
[05:04:13] <Qten> and they use zfs filers
[05:04:17] <richlowe> coraid use AoE _and_ plan9
[05:04:19] <richlowe> draw your own conclusions.
[05:04:31] <wesolows> yes, we consider coraid entertainment.
[05:04:55] *** artimus has quit IRC
[05:05:04] <Qten> well now i just have to ask why!
[05:05:25] <Qten> they seem to be set on selling a service using it even nexenta support aoe coraid chassis
[05:05:31] <wesolows> the problem here is twofold: first, ZFS expects that its devices are generally available.  Failure of a device in terms of it just plain going away is tolerated, but it's not designed to have large numbers of devices coming and going, getting slow and then fast again, etc.
[05:06:13] <wesolows> second, the complexity of managing a distributed system requires that the tooling and infrastructure be designed for such a thing.  ZFS isn't designed for that, at all.
[05:06:47] <Qten> is the key hear the large number of devices? i would have thought the slow and fast would have been a pool would only be as fast as its slowest disk?
[05:06:50] <wesolows> I could also go into all the human error inherent in managing huge quantities of iSCSI targets, servers with iSCSI target services, etc.
[05:07:27] <wesolows> Tell you what.  Go try it, run it in production for a year with 500 servers, and then I won't need to tell you a thing.
[05:07:38] <wesolows> Some people have to learn the hard way.
[05:07:47] <ira> wesolows: s/500/50/
[05:07:51] <Qten> i'm asking questions!
[05:07:54] <wesolows> ira: I'd lose my mind at 5.
[05:08:29] <ira> wesolows: But someone MIGHT pull off 5.
[05:08:29] <wesolows> The number of devices is a big problem, but not at all the only one.
[05:08:30] <wesolows> I've explained many problems so far.
[05:08:40] *** Webhostbudd has joined #smartos
[05:08:48] <Qten> i do see your points
[05:08:57] <Qten> and admittedly its overly complex in hindsight
[05:09:07] <wesolows> as for why ATA over Ethernet sucks, I'll direct your attention to the two basic components underlying it...
[05:09:32] *** daleg has quit IRC
[05:09:51] <wesolows> ATA.  Bastard stepchild of block protocols.  Inferior to SCSI in every conceivable way.  Evolved, not designed, from ancient PC hardware intended primarily to be low-cost.
[05:10:27] <wesolows> Ethernet.  Not a storage transport.  Designed for small-scale LANs with small packets for low-latency low-throughput applications.
[05:10:49] <wesolows> of the two, ATA is far weaker.  iSCSI at least sort of works, sometimes.
[05:11:20] <ipAlreadyTaken> IMO the biggest argument against these types of storage setups is just the added points of failure. Sure you can and levels or redundancy, but then you are just adding more places for failure. Take a set of disk on a local machine, sure the disk can go bad, os panic ext, BUT lets say you add some sort of network storage, now you have to worry about not only your local machine but the network switch, the remote disks, remote software, and everythin
[05:11:27] <wesolows> Plan9... this is where I fail to keep a straight face.
[05:11:38] <ira> wesolows: I thought ATA over Ethernet didn't even bother with TCP though ;)
[05:11:54] <wesolows> ipAlreadyTaken: exactly.  There are a billion little things that can go wrong, and most of them expose a failure mode to ZFS that it is not designed to handle.
[05:12:21] <wesolows> ira: right.  Hence, not routable.  In that sense, I suppose it limits the scale of your misadventure.
[05:12:52] <ira> wesolows: I guess… never underestimate the power of stupidity ;)
[05:12:58] *** jamesd_ has joined #smartos
[05:13:04] <wesolows> ipAlreadyTaken: This is also why "HA" clustering typically has more downtime than supposedly SPOF storage.  See the various github outages.
[05:13:10] <ipAlreadyTaken> Wesolows only to how many switches you daisy chain together :)
[05:13:34] *** jdefelice has joined #smartos
[05:14:05] <Qten> so how would you obtain a large amount of HA storage ? using 2 servers a jbod and SAS?
[05:14:32] <ira> Qten: I'd ask, why do you want it?
[05:14:38] *** xinkeT has quit IRC
[05:14:48] <wesolows> Qten isn't a troll; he's the narrator in my "this is your life" montage.
[05:14:58] <wesolows> we're in 2006 now.
[05:15:04] <Qten> local storage can only grow so far
[05:15:06] *** xinkeT has joined #smartos
[05:15:13] <Qten> fantastic thanks for the taunts
[05:15:14] <wesolows> what's your actual use case?
[05:15:23] <wesolows> let's get serious here.
[05:15:38] <Qten> several compute nodes with local storage 1 needs 100TB
[05:15:47] *** jamesd has quit IRC
[05:15:47] <Qten> for a customer for example
[05:15:58] <wesolows> no examples.  what's YOUR use case?
[05:16:10] <Qten> that is my use case
[05:16:10] *** Tekni has joined #smartos
[05:16:12] <ira> 100TB?  You mean a 4/5 U DAS?
[05:16:27] <wesolows> Then use a 4U 36x3.5" chassis with a simple 2S board and stick 4TB drives in it.
[05:16:30] <wesolows> Use RAIDZ2
[05:16:50] <wesolows> Problem solved.
[05:16:52] <Qten> yeah that could give you it on 1 server however now you need to replicate it
[05:17:04] <wesolows> Shard.
[05:17:13] <Qten> or shared yes
[05:17:21] <wesolows> No, shard.  Not shared.
[05:17:36] <Qten> zfs snapshot replication?
[05:17:36] <wesolows> redundancy is properly handled at the application level.
[05:17:48] *** xinkeT has quit IRC
[05:17:55] <wesolows> why does this need to be replicated?
[05:18:26] <wesolows> you have raidz2 with 3 hot spares; your data's not going away.  If something in the system fails, you replace it.
[05:18:27] <Qten> fire, disaster etc
[05:18:36] <wesolows> database replication, then.
[05:18:37] *** bpaluch_ has joined #smartos
[05:18:59] <wesolows> or you can use delegated datasets in SmartOS and do ZFS snapshots and send/recv if you wish.
[05:19:09] <wesolows> It depends what you're storing...
[05:19:09] <Qten> true
[05:19:16] *** iyp has quit IRC
[05:19:23] *** bpaluch has quit IRC
[05:19:26] <wesolows> The reality is that DC fires don't happen.
[05:19:33] *** marsell has joined #smartos
[05:19:40] <wesolows> But if you're running a bank or whatever, sure, use replication.
[05:19:40] <Qten> also true
[05:20:08] <wesolows> If you're just some $500k a year retail site or whatever, it's not worth the money.
[05:20:23] <Qten> see your point
[05:22:22] <ipAlreadyTaken> Qten: let's look at how this would be with iscsi: if you needed 100tb on a single vol then it's the same number of disk, and if you want the security of a second system you will need another system that has another 30 disks. So no matter what you have about 8u of disk and 60 disk. If you make the storage network based you are adding switches and more servers for the actual processing. If you want that replication setup either way you have the same
[05:22:23] <wesolows> It never gets old watching people bend over backwards to install mindblowingly complicated "HA" systems and then they use shitty SATA disks or rotten protocols like AoE or bug-riddled software on top of it and then wonder why it's not 17 9s.
[05:23:24] <wesolows> The #1 most important place to improve reliability is your application software.  The #2 place to improve reliability is by choosing hardware carefully, with a keen eye toward reducing the amount and complexity of firmware involved.
[05:23:54] <wesolows> The next place to invest is in networking redundancy.  Most people who nail those three will beat any commercial clustering solution in availability.  Yes, really.
[05:25:18] *** abhishekvasisht has joined #smartos
[05:25:34] <ira> wesolows: People forget a modern server is so redundant internally anyways.  Power supplies, drives, controllers...etc.
[05:27:22] <Qten> its a refreshing take i'll say that
[05:27:32] <wesolows> ira: yeah, exactly.
[05:27:45] <wesolows> You have to match the risks against the costs.
[05:28:04] <Qten> i personally like the idea of local storage as when a HA Cluster goes tits up everyones offline as apposed to 1 server with a few clients on it
[05:28:11] <wesolows> If you're doing $100B a year on the box, fine, invest in redundancy.  It probably won't work, but at least you won't get fired when it fails.
[05:28:23] <ira> Qten: Or runs out of bandwidth etc...
[05:28:25] <rmustacc> An interesting thing to remember though is HA clusters are generally about minimizing downtime, not being the most available.
[05:28:31] <wesolows> Qten: exactly
[05:30:10] *** tonyarkles has joined #smartos
[05:30:31] <wesolows> That's the thing -- if you're going to do something complex, make sure it fails one bit at a time.
[05:31:16] <wesolows> "HA" clustering never does.  It fails catastrophically.
[05:32:34] <Qten> agreed
[05:33:21] <wesolows> Oh, and I shouldn't have, but I left off #0 earlier... operational runbook.  Almost all outages are caused by operator error.
[05:33:45] <Qten> yeah we've had a few of them ourselves heh
[05:34:04] <ipAlreadyTaken> Like every aws outage in the past 2 yrs
[05:34:38] <wesolows> AWS outages are mostly caused by EBS cascading failure.
[05:35:17] <wesolows> just like github outages are mostly caused by cascading failure when HA storage goes on the blink.
[05:35:21] <wesolows> There's a common thread there...
[05:35:50] <richlowe> "things that make Keith angry"
[05:36:10] <wesolows> And of outages that aren't triggered directly by operator error, I'd wager at least 1/3 start with "... we had our vendor techs on site to perform a firmware upgrade on the HA $thing and ..."
[05:36:23] <Qten> on a semi related note with zfs does a zill help turn random writes into sequential writes? or is it just a method of quickly updating metadata so it can return a sync operation and the writes are still random?
[05:36:58] <wesolows> it is a log, like any other log.  it allows sequential writing of intent records.
[05:37:17] <wesolows> its benefits are maximised when the zil is located on a fast slog, typically an SSD.
[05:37:48] <Qten> so what seeminly would have been for example 2000iops write into 500 iops on disk?
[05:38:04] <Qten> 2000iops random
[05:38:09] <wesolows> possibly.  it blows out the window for aggregating and scheduling I/O.
[05:39:05] <Qten> so what your basically saying is if you "need" 2000 iops you need 2000 disk iops as well
[05:39:14] <wesolows> no
[05:39:38] <wesolows> if you need to do 2000 small-block synchronous writes per second, you need a slog that can sustain > 2000 such writes.
[05:39:47] <wesolows> ideally more.
[05:40:57] <Qten> agreed, however i'm more so saying if i need to do 2000 small-block synchronous writes per second a SSD isnt going to help me reduce that into spinning disk iops
[05:41:08] <wesolows> sure it will
[05:41:30] <wesolows> because when we write out, we'll be able to do 112k or 128k writes instead of 512 byte writes
[05:41:50] <ira> "It does all the time here."
[05:42:04] <wesolows> this is all about block allocation, I/O aggregation, and I/O scheduling.
[05:42:10] <Qten> i guess i'm trying to wrap my head around it like a teird storage equallogic were it will commit all writes to ssds and move them down to SAS as the need rises
[05:42:20] <wesolows> that's not how it works.  at all.
[05:42:27] <ira> qten: This isn't EQ.
[05:42:31] <wesolows> tiered storage is the opposite of hybrid storage.
[05:43:40] <Qten> ira: also agreed just looking at different tech
[05:43:42] <wesolows> writes are committed to the slog immediately, provided you can allocate space on it and so on
[05:43:53] <ira> Qten: There's 2 ways to look at things, every block of space is sacred, because we payed $$$$ for it.  Or, disks are tools to use for various purposes.  The latter is the attitude of ZFS.  Got a SSD?  Let's use it as a write cache, or a read cache.
[05:44:05] <wesolows> the slog is never read again unless power is lost or the system panics
[05:44:42] <ira> Not actually allocate the block there, per-say… (The long term storage isn't on the slog/L2ARC.)
[05:44:45] <wesolows> in the meantime, the presence of the data on persistent storage allows ZFS to ack the write, and gives it time to rearrange the blocks of data in DRAM into large I/Os to the main pool devices.
[05:45:12] <wesolows> ira: well, sure.  but if you fill your slog, you're going to the mail pool.  Imagine a 50 MB slog...
[05:45:16] <ira> (and wesolows will do a better job explaining it than me :) )
[05:45:46] <ira> wesolows: I don't think I've used under a 32GB one… Considered as low as 4-8GB.
[05:45:48] <wesolows> there are plenty of white papers on this, I hope, on the Internet
[05:45:49] *** tonyarkles has quit IRC
[05:45:59] <wesolows> ira: 8 is fine.  Not sure about 4 but probably mostly fine too.
[05:46:07] *** szaydel has joined #smartos
[05:46:15] <wesolows> I've heard of a few customers getting past 8, but it's likely very rare.
[05:46:20] <wesolows> (not our customers, mind)
[05:46:33] <Qten> wesolows: fair enough pretty much confirms most of which i've read
[05:47:35] <ira> 8… You'd have to almost tune to hit 8...
[05:50:06] <ipAlreadyTaken> 2 Hous and 12 minutes later platform installation is complete is complete! I know what I am buying first thing tmrw morning!!
[05:50:22] <ira> ipAlreadyTaken: A good USB stick?
[05:50:49] <ipAlreadyTaken> Ya think?
[05:50:51] <wesolows> ouch.
[05:51:16] *** nefilim has quit IRC
[05:51:31] <ira> ipAlreadyTaken: Been there, done that… though not THAT badly.
[05:51:53] <ipAlreadyTaken> I was talking to a guy at ddrdrive a few months ago and he said they had a customer with 16BG on one of their cards
[05:52:50] <ipAlreadyTaken> But he didn't know what actual usage was
[05:54:00] *** szaydel has quit IRC
[05:54:33] <ira> at 168G… I begin to wonder if it shouldn't be in main memory… eep.
[05:54:43] <wesolows> I've heard something similar from STEC, customers asking for a bigger ZeusRAM.
[05:55:09] <wesolows> oh, 168... that's crazy.
[05:55:38] <ira> I'd just like a zeusram in a 2.5in form factor...
[05:55:46] <wesolows> if you're filling 168 GB in the txg sync timeout, you must have one awesome main pool!
[05:55:47] <sheppard> booo ddrdrive is full height
[05:55:50] <wesolows> ira: we all would
[05:55:54] <sheppard> can't use it in my chassis =\
[05:55:59] <Qten> whats the best price have you guys found for the ZeusRAM drives?
[05:56:13] <wesolows> I can't really say.  Like all things, it depends on volume.
[05:56:22] <Qten> true
[05:56:39] <Qten> 2500 just seems excessive
[05:56:53] <wesolows> you can probably find 1 between 2k and 2500.
[05:56:56] <ira> Is it?  For what it does in the architecture?  Meh.
[05:57:02] <Qten> true
[05:57:13] <Qten> however if its per local box yeah
[05:57:23] <ira> Go get a flash cache card from a vendor… :)
[05:58:59] <ipAlreadyTaken> How do the Zeusram drives deal with a power outage?
[05:59:12] <wesolows> by splatting DRAM into flash
[05:59:19] <ira> Caps, and write back to nvram as I remember.
[05:59:24] <wesolows> powered by supercaps
[06:00:38] <ipAlreadyTaken> Oh ok. That's the thing that always scared me about getting the ddrdrive a was the power failure. They use an external power source (either a batter or 110), but it never sat right with me
[06:02:14] <wesolows> I prefer the caps myself, but Li-ion can store a lot more energy per unit volume.  Hardware is hard...
[06:04:17] <ipAlreadyTaken> The bigger issue I think I had was that it was just plugged into the backside of the card and sat outside the chassis. I could see somebody unplugging it "for just one second" and then walking away. Stuff like that scares me
[06:04:39] *** ira has quit IRC
[06:15:43] *** sachinsharma has joined #smartos
[06:18:15] *** MTecknology has left #smartos
[06:25:45] *** echelog-1 has joined #smartos
[06:26:25] *** ipAlreadyTaken has quit IRC
[06:27:59] *** wolfeidau has quit IRC
[06:31:07] *** darjeeling has quit IRC
[06:43:41] *** darjeeling has joined #smartos
[06:46:10] *** ipAlreadyTaken has joined #smartos
[06:52:55] *** Qten has quit IRC
[06:54:00] *** Qten has joined #smartos
[06:58:47] *** ipAlreadyTaken has quit IRC
[07:01:40] *** sachinsharma has quit IRC
[07:02:26] *** sachinsharma has joined #smartos
[07:12:23] *** nefilim has joined #smartos
[07:25:40] *** wolfeidau has joined #smartos
[07:25:57] *** sachinsharma_ has joined #smartos
[07:26:17] *** Qten has quit IRC
[07:26:58] *** sachinsharma has quit IRC
[07:26:58] *** sachinsharma_ is now known as sachinsharma
[07:27:50] *** Qten has joined #smartos
[07:42:10] *** miller71 has quit IRC
[08:11:37] *** nefilim has quit IRC
[08:25:33] *** alucardX has joined #smartos
[08:28:51] *** darjeeling has quit IRC
[08:41:12] *** Cpt-Oblivious has quit IRC
[08:56:19] *** Andyy has quit IRC
[09:01:21] *** ipalreadytaken has joined #smartos
[09:11:49] *** xinkeT has joined #smartos
[09:14:12] *** sachinsharma_ has joined #smartos
[09:14:16] *** sachinsharma has quit IRC
[09:14:16] *** sachinsharma_ is now known as sachinsharma
[09:14:36] *** Teknix has quit IRC
[09:24:46] *** texarcana has quit IRC
[09:26:00] *** texarcana has joined #smartos
[09:26:04] *** bens1 has joined #smartos
[09:31:33] *** marsell has quit IRC
[09:56:33] *** ipalreadytaken has quit IRC
[09:56:41] *** mg576 has joined #smartos
[09:57:22] *** ipalreadytaken has joined #smartos
[10:01:30] *** ipalreadytaken has quit IRC
[10:02:10] *** ipalreadytaken has joined #smartos
[10:04:14] *** Andyy has joined #smartos
[10:04:17] *** miller7 has joined #smartos
[10:06:17] *** ipalreadytaken has quit IRC
[10:09:11] *** wolfeida_ has joined #smartos
[10:10:48] *** wolfeid__ has joined #smartos
[10:11:25] *** wolfeidau has quit IRC
[10:13:47] *** wolfeida_ has quit IRC
[10:14:32] *** miller71 has joined #smartos
[10:15:12] *** miller7 has quit IRC
[10:15:55] <Andyy> Hi all :)
[10:18:11] *** eliasgs has joined #smartos
[10:18:29] *** KermitTheFragger has joined #smartos
[10:19:52] <Andyy> I wonder if smartos can do kvm live migration of running guests
[10:20:38] <mg576> Andyy: smartos cannot do this afaik, sdc might be able but i dont think either
[10:20:59] <jperkin> no.
[10:21:31] <Andyy> mg576: I'm trying to get in touch with joyent sales as I'd like to have a go on sdc too.
[10:21:40] <mg576> there is a vmadm send option to send machines to another machine but its rather new i am experimenting with it
[10:22:56] <mg576> Andyy: jperkin is an engineer at joyent so thats a definite answer :) just surprised to see him here so early (or late) since i thought joyent is on the west-coast
[10:23:19] <jperkin> I'm in the UK
[10:23:40] <Andyy> ouch. hi jperkin.
[10:23:50] <mg576> jperkin: yeah just looked up your profile ... worked at bbc !
[10:24:27] <mg576> good to know some knowledgable people about smartos live close to my timezone (germany)
[10:24:42] <Andyy> <- Italy :)
[10:25:08] <Andyy> I could drive to your offices mg576... by car.
[10:25:50] <Andyy> jperkin: are there plans to support qemu/kvm migration under smartos ?
[10:26:15] <mg576> Andyy: i am not sure this would really solve the problems smartos users have because the kind of apps we write are multi machine type apps
[10:26:37] <mg576> Andyy: in my experience, live migration is relevant for old style server apps
[10:26:43] <jperkin> I don't know of any, it's more of a question for sales as you say
[10:26:45] <mg576> Andyy: which you might have of course
[10:27:26] <Andyy> I agree... I'm trying to understand if it's viable solution/alternative for cloud infrastructure. You'll normally end up with many customer machines that you cannot reboot at will.
[10:28:19] <Andyy> I mean an alternative to openstack/vmware clusters with shared storage.
[10:28:29] <mg576> Andyy: right. but that depends on the service model you promise them. the hosting partners we have always reboot our machine whenever they want. so does amazon BTW (sometime machines actually vanish for good)
[10:29:01] <mg576> Andyy: joyent engineers (correct my when i am wrong) are not big fans of shared storage for vms for a number of reasons
[10:29:09] <mg576> and i am not either anymore
[10:29:38] <Andyy> I am also not a fan of shared storage for many reasons.
[10:29:41] <mg576> at least the os partition should be locally stored. additional storage can be integrated of course (iscsi etc)
[10:30:28] <mg576> right now we are looking into this to build up a new LOB cloud server application where each customer would get his own OS zone
[10:31:02] <mg576> that way there is almost no process overhead, huge sharing benefits and great security
[10:31:55] <mg576> so we dont need kvm itself a lot
[10:32:45] *** ipalreadytaken has joined #smartos
[10:32:48] <Andyy> I understand. smartos is perfect if you can control all your instances. I'm planning to move almost all our our hosting business to smartos zones
[10:33:09] <Andyy> We currently run off openvz containers under linux.
[10:33:20] <Andyy> before that it was vserver and freevsd earlier.
[10:33:40] <mg576> btw .. the newsgroup is quite helpful -- even when they are in their infancy documentation wise
[10:34:02] <Andyy> newsgroup?
[10:34:41] <Andyy> freeVsd... that's correct.
[10:35:17] *** mgt576 has joined #smartos
[10:36:10] <mgt576> wreird ... my nick just changed to my original name (instead of the alt)
[10:37:01] <Andyy> happened to me a few times, until I decided to use Andyy as the nick. I think it's because your nick might be registered.
[10:37:35] <mgt576> i think there was an old connection active (i actually registered my nick and am identified but ok)
[10:38:29] <Andyy> did you refer to "gmane.os.smartos.general"
[10:38:33] <Andyy> ?
[10:38:54] <mgt576> btw driving distance looks like 8 hours :)
[10:39:06] *** mg576 has quit IRC
[10:39:42] <Andyy> mgt576 yes... very close.
[10:41:07] *** ipalreadytaken has quit IRC
[10:42:00] <mgt576> Andyy: well i am not using gmane (and it does not look updated anyway) i just email to smartos-dicuss at lists dot smartos.org
[10:44:08] <Andyy> thanks.
[10:48:04] *** khushildep has joined #smartos
[11:00:53] *** marsell has joined #smartos
[11:01:00] *** Tekni has quit IRC
[11:01:17] *** Tekni has joined #smartos
[11:09:40] *** jwk404_ has joined #smartos
[11:10:14] *** jwk404 has quit IRC
[11:10:16] *** pinja has quit IRC
[11:10:22] *** wesolows has quit IRC
[11:10:23] *** Tekni has quit IRC
[11:10:29] *** Tekni has joined #smartos
[11:13:19] *** pinja has joined #smartos
[11:13:20] *** wesolows has joined #smartos
[11:19:01] *** Tekni has quit IRC
[11:21:23] *** Teknix has joined #smartos
[11:30:04] <estibi_remote> I don't have an access to the internet but I can download dataset image (imgadm)  and save it locally. Where should I save it on a SmartOS?
[11:31:14] <jperkin> anywhere, then do a manual imgadm install along with the manifest
[11:33:28] *** eliasgs has quit IRC
[11:36:38] *** Teknix has quit IRC
[11:38:04] *** Teknix has joined #smartos
[11:44:39] *** Teknix has quit IRC
[11:46:04] *** Teknix has joined #smartos
[11:49:51] *** miller71 has quit IRC
[11:50:02] *** miller7 has joined #smartos
[11:51:48] <miller7> Hey guys, what's the best way to delete my disks and reboot smartos in order to re-create everything?
[11:52:51] <jperkin> boot into the noinstall option, dd /dev/zero over the first few MB of each disk and reboot?
[11:54:27] <mgt576> miller7: we just rebooted with noinstall and detroyed the pool...worked fine
[11:55:19] <miller7> thanks
[12:03:37] *** Teknix has quit IRC
[12:05:02] *** Teknix has joined #smartos
[12:10:30] *** Teknix has quit IRC
[12:11:54] *** Teknix has joined #smartos
[12:16:28] *** abhishekvasisht has quit IRC
[12:27:32] *** Teknix has quit IRC
[12:29:01] *** Teknix has joined #smartos
[12:39:06] *** bens1 has quit IRC
[12:44:30] *** texarcana has quit IRC
[12:48:33] *** Teknix has quit IRC
[12:49:05] *** sachinsharma has quit IRC
[12:49:57] *** Teknix has joined #smartos
[12:57:26] *** leecallen has quit IRC
[12:58:06] *** leecallen has joined #smartos
[13:01:55] *** jamesd_ has quit IRC
[13:02:08] *** jamesd has joined #smartos
[13:02:39] *** Teknix has quit IRC
[13:04:04] *** Teknix has joined #smartos
[13:06:06] *** khushildep has quit IRC
[13:31:28] *** Teknix has quit IRC
[13:32:53] *** Teknix has joined #smartos
[13:33:08] *** ira has joined #smartos
[13:39:02] *** ipalreadytaken has joined #smartos
[13:39:46] *** ipalreadytaken has joined #smartos
[13:44:36] *** ipalreadytaken has quit IRC
[13:45:44] *** enmand has joined #smartos
[13:47:33] *** Teknix has quit IRC
[13:48:21] *** alucardX has quit IRC
[13:48:58] *** Teknix has joined #smartos
[14:01:48] <nahamu> jperkin: any ETA on when you'll have a chance to look at those spice packages?
[14:05:07] *** Sachiru has joined #smartos
[14:06:35] *** Sachiru has quit IRC
[14:06:59] *** Sachiru has joined #smartos
[14:12:47] *** simong has joined #smartos
[14:17:20] *** sachinsharma has joined #smartos
[14:20:44] *** sachinsharma has quit IRC
[14:26:22] *** Teknix has quit IRC
[14:28:05] *** Teknix has joined #smartos
[14:33:04] *** artimus has joined #smartos
[14:37:08] <nahamu> rmustacc: whenever you wake up: https://github.com/nshalman/qemu/commits/qemu-kvm-1.1.2-for-illumos
[14:37:26] <nahamu> have to add --disable-kvm-device-assignment when you run configure
[14:37:47] <Alasdairrr> whats this?
[14:37:48] <nahamu> but it passed a quick smoke test last night of booting a spice enabled VM...
[14:38:06] <nahamu> I merged in the qemu-kvm 1.1.2 tag
[14:38:27] <nahamu> my qemu tree was originally based on qemu-1.1.2
[14:38:34] <Alasdairrr> ah
[14:39:20] <nahamu> it might get us some extra performance
[14:43:54] *** ajeffco has joined #smartos
[14:53:11] <nahamu> uhhh... holy crap... that might have actually boosted network performance.
[14:53:59] <MerlinDMC> nahamu, that (dunno how it's called) host-to-guest filesharing of qemu is not functional or?
[14:54:17] <nahamu> MerlinDMC: I've never tried to use it.
[14:54:25] <nahamu> were you trying?
[14:54:40] <MerlinDMC> 9p was it i think
[14:54:52] <MerlinDMC> nah ... don't have your updated qemu ^^
[14:55:19] <nahamu> I need to poke it a bit more, but I'm very tempted to push this one out to my image server
[14:56:07] <nahamu> my production VM (qemu-1.1.2) iperf client run got ~230 Mbits/sec
[14:56:13] <MerlinDMC> but i gues also the guest has to have components to mount 9p shares ... hmm
[14:56:36] <nahamu> a testing VM on my dev box with this new qemu-kvm-1.1.2 got ~850 Mbits/sec
[14:57:00] <nahamu> I need to verify that the old qemu on the dev box is still slower.
[14:57:45] <nahamu> but this is exactly what rmustacc was describing when he told me I should be using qemu-kvm rather than vanilla qemu
[14:58:24] <MerlinDMC> maybe that will close the network performance thread in the mailing list as well ^^
[14:58:36] <MerlinDMC> you should check for the package loss they describe as well
[15:02:07] <miller7> Is it normal for SmartOS to take ++ minutes to load if on dhcp and cannot get an IP?
[15:02:28] <MerlinDMC> miller7, the platform or a zone?
[15:03:21] <miller7> the platform (host OS), booting from CD. I made the mistake to select the wrong NIC (one is connected in public IP and the internal on DHCP). I chose the public NIC and set to dhcp and it takes ages to give me login prompt
[15:04:23] <jperkin> nahamu: sorry, I was going to do it, but got sidetracked by the fact github merge wanted to use my default non-joyent email address, and I started to look at cmdline ways to do it, and then got side-tracked :)
[15:04:41] <nahamu> jperkin: no worries!
[15:04:44] *** khushildep has joined #smartos
[15:04:51] *** bayoda has quit IRC
[15:05:04] <jperkin> it's not normally a big deal, but some of our scripts rely on parsing @joyent.com authors for fixing up merge conflicts.
[15:05:18] <nahamu> ah
[15:05:29] <jperkin> if you dump them in a tar somewhere I'll probably just git add them by hand :)
[15:06:08] <MerlinDMC> miller7, it might wait for ntp updates ... but can't tell for sure ... i'll see if i can verify that here and it is the network connectivity
[15:06:13] <nahamu> MerlinDMC: my excitement was too soon. old qemu on dev machine got ~850Mbits/sec too...
[15:06:21] <nahamu> must be a network layout difference.
[15:06:32] <nahamu> must be a network layout difference.
[15:06:54] <nahamu> jperkin: I'll post a tarball in a few minutes.
[15:10:26] *** d[^_^]b has quit IRC
[15:10:34] *** d[^_^]b has joined #smartos
[15:10:35] <miller7> MerlinDMC: could it be a DNS / hostname issue? Any pointers in that direction? My hostname is something like bc-ae-cx-xx-xx-xx or similar
[15:11:36] <nahamu> jperkin: http://www.shalman.org/spice/pkgsrc_joyent_spice-stuff.tar.gz
[15:12:23] <nahamu> the tarball includes the pkgsrc/joyent/ part of the path
[15:14:10] <miller7> also, I see this: "multipath status: degraded: path 2 mpt_sas2/disk@" for both disks I use for test. Is it something I need to do?
[15:17:24] *** Sachiru has quit IRC
[15:18:21] *** enmand has quit IRC
[15:18:52] <jesse_> jperkin, do pkgsrc-multiarch bugs go to smartos-live/issues?
[15:19:06] <jperkin> jesse_: no, pkgsrc/issues
[15:22:11] <jesse_> jperkin, https://github.com/joyent/pkgsrc/issues/33
[15:22:21] <jesse_> just wanted to document that if others come across it=)
[15:22:41] <jperkin> thanks, interesting..
[15:22:43] <jesse_> spent over a day staring at that problem
[15:23:25] <jesse_> just cflags+=-D_REENTRANT and rebuild and it works
[15:23:32] *** jdefelice has quit IRC
[15:23:55] <jesse_> (or might have been CPPFLAGS, I didn't do clean before CFLAGS and it wasn't enough)
[15:25:25] <jesse_> might actually be a problem in other pkgsrc packages, too
[15:25:36] <jesse_> (pkgsrc sets?)
[15:25:47] <jperkin> packages
[15:26:07] <jperkin> nahamu: https://github.com/joyent/pkgsrc-joyent/commit/e91a360787bc1173962cdeb57e258f7cf0c71f9d - I'll kick off a multiarch build with them in.
[15:26:08] <jesse_> I mean, '32bit set'/'64bit set'/'illumos set'
[15:26:45] <nahamu> jperkin: thanks!!
[15:26:56] <nahamu> I imagine there will be a nit somewhere
[15:27:33] <nahamu> but I think I got pretty close.
[15:27:53] <jperkin> oh, libxcb already has some -D_REENTRANT but it's conditionalised, will look further..
[15:31:08] <jperkin> right, someone typo'd CFLAGS vs CPPFLAGS
[15:31:13] *** jdefelice has joined #smartos
[15:31:57] *** enmand has joined #smartos
[15:32:14] *** bens1 has joined #smartos
[15:33:10] *** bayoda has joined #smartos
[15:37:15] *** CarlosC has joined #smartos
[15:37:52] *** jdefelice has quit IRC
[15:38:12] *** jdefelice has joined #smartos
[15:39:37] *** jdefelice has quit IRC
[15:40:05] *** jdefelice has joined #smartos
[15:40:25] <nahamu> MerlinDMC: I've pushed a binary image of qemu-kvm-1.1.2 to datasets.shalman.org if you want to try it out.
[15:40:39] <jperkin> jesse_: I fixed it upstream.
[15:40:42] *** ipalreadytaken has joined #smartos
[15:41:45] *** jdefelice has quit IRC
[15:42:03] *** jdefelice has joined #smartos
[15:44:39] *** jdefelice has quit IRC
[15:46:02] *** jdefelice has joined #smartos
[15:49:07] *** sachinsharma has joined #smartos
[15:51:31] *** iyp has joined #smartos
[15:52:13] <miller7> My SmartOS booting times: showing dots for 7 minutes. Then showing "SunOS Release 5.11" for 15 minutes! I am sure this is not normal... What could I do to see what the problem is? This is a machine with 24 cores and 96GB of RAM, it's not an 8086...
[15:52:55] <nahamu> miller7: the first part probably has to do with your BIOS and your USB stick.
[15:53:12] <nahamu> the latter you might be able to get more information by telling the kernel to boot verbosely
[15:54:07] <miller7> nahamu: I load from a CD-ROM and tried to check everything I could on my BIOS (googling and reading the mailing lists)
[15:54:22] <miller7> nahamu: How can I tell the kernel to boot verbosely?
[15:55:02] <nahamu> miller7: hang on, I forget... let me check
[15:56:23] <nahamu> try adding "-vd -m verbose" to the kernel command line
[15:56:54] *** autif has joined #smartos
[15:56:58] <miller7> thank you very much
[15:57:41] <jesse_> jperkin, oh wow, it went so far up=)
[16:00:56] *** tonyarkles has joined #smartos
[16:01:51] *** iyp has quit IRC
[16:02:23] *** mg576 has joined #smartos
[16:02:46] *** khushildep has quit IRC
[16:04:52] *** mgt576 has quit IRC
[16:05:05] *** khushildep has joined #smartos
[16:06:44] *** des2 has joined #smartos
[16:07:03] *** khushildep has quit IRC
[16:07:23] *** khushildep has joined #smartos
[16:08:29] *** sachinsharma has quit IRC
[16:10:15] *** jamesd_ has joined #smartos
[16:13:08] *** jamesd has quit IRC
[16:13:28] *** daleg has joined #smartos
[16:15:11] *** daleg has quit IRC
[16:15:32] *** daleg has joined #smartos
[16:16:13] <jesse_> jperkin, how are the X fonts supposed to be installed? font-adobe-* etc. separately or does the modular-xorg-fonts have a sane set/all?
[16:16:33] <jperkin> I think that's a bit of a mess, and they are separate
[16:16:35] <jesse_> and as dependencies for a package, which should it depend on if it names fonts?
[16:16:46] <autif> I am very new user of the JPC - Should I be powering off the VM when I do not plan to use it (say over the weekend)? http://wiki.smartos.org/display/DOC/Zones does not say anything regarding that. And as you can see - I am a newbie :-)
[16:17:32] <jperkin> autif: more of a #joyent question, but you are charged for machines which are powered off but still exist, you have to destroy the machine before it will stop being billed.
[16:18:03] *** iyp has joined #smartos
[16:18:21] <autif> jperkin: thanks!
[16:18:28] <jesse_> jperkin, did you mean 'they are separate packages' or 'you need to install them separately'?
[16:19:38] <jperkin> jesse_: modular-xorg-fonts will install a bunch but I think there are still packages which require certain fonts which aren't depended upon correctly.
[16:20:02] <jesse_> ah, ok
[16:20:20] <jesse_> so I should note which fonts I refer to so that I'll get the dependencies right later
[16:20:31] <jperkin> ideally there should be fine-grained dependencies for each package so they only get what they need.
[16:20:42] <jperkin> but I imagine that's going to be pretty tricky
[16:20:59] <jesse_> yes
[16:22:03] *** neophenix has joined #smartos
[16:22:21] <jesse_> and dejavu-ttf is not even named font-dejavu-ttf
[16:23:14] *** simong has quit IRC
[16:23:57] *** tonyarkles has quit IRC
[16:25:05] <miller7> nahamu: I did the verbose kernel boot and it stuck a very long time after a line saying something like pci@0,0 blah blah (mpt_sas3) online
[16:25:34] <miller7> then next line showed (after the delay) was: dump on /dev/zvol/disk/zones/dump size 4096 MB
[16:25:43] <miller7> between those two lines it took ages
[16:26:10] <nahamu> so that's way out of my depth... could be an issue with the mpt sas driver...
[16:26:47] <nahamu> in a few hours when the Joyent guys wake up and come online they might be able to help diagnose further.
[16:26:57] <miller7> nahamu: Well, thanks for helping me so far anyway. It is very much appreciated!
[16:27:33] <MerlinDMC> nahamu, thx ... my dataset mirror already found that and synced it ;)
[16:27:49] *** tonyarkles has joined #smartos
[16:27:50] *** ipalreadytaken has quit IRC
[16:28:13] *** ipalreadytaken has joined #smartos
[16:29:01] <jesse_> illumost/smartos doesn't have any fonts from sun?
[16:29:07] <nahamu> MerlinDMC: so the bits are already on your side of the pond... even less excuse for you not to try them out... :-P
[16:29:33] <MerlinDMC> yeah ... have to push some data first
[16:50:21] *** khushildep has quit IRC
[16:51:41] *** enmand has quit IRC
[16:54:11] *** khushildep has joined #smartos
[16:54:52] *** khushildep has quit IRC
[16:55:23] *** khushildep has joined #smartos
[16:56:24] *** enmand has joined #smartos
[17:01:05] *** enmand has quit IRC
[17:01:24] *** enmand has joined #smartos
[17:03:53] *** bayoda has quit IRC
[17:05:39] <Licenser> dun dun dun
[17:05:51] *** khushildep has quit IRC
[17:08:24] *** ajeffco has quit IRC
[17:14:26] <nahamu> greetings, Licenser
[17:14:34] <Licenser> :) how is life?
[17:14:41] <nahamu> it's good.
[17:14:51] <Licenser> huzza!
[17:20:51] *** bayoda has joined #smartos
[17:22:58] *** dap has joined #smartos
[17:25:26] *** d[^_^]b has quit IRC
[17:31:35] *** ipalreadytaken has quit IRC
[17:32:43] *** d[^_^]b has joined #smartos
[17:33:56] *** sachinsharma has joined #smartos
[17:34:50] *** bayoda has quit IRC
[17:39:08] *** jdefelice has quit IRC
[17:39:37] *** jdefelice has joined #smartos
[17:41:40] *** jdefelice has quit IRC
[17:42:07] *** quasi_ is now known as quasi
[17:42:13] *** jdefelice has joined #smartos
[17:42:33] *** nefilim has joined #smartos
[17:42:36] *** quasi has quit IRC
[17:42:36] *** quasi has joined #smartos
[17:43:39] *** bayoda has joined #smartos
[17:43:44] *** jim80net has joined #smartos
[17:48:56] *** serzhit has joined #smartos
[17:51:00] <nahamu> jperkin: the cleanup you had to do wasn't as bad as I'd feared.
[17:51:18] <serzhit> HEllo. Could anybody help. I installed katest smartos with one physical vnic. Then I create virtual ,achine with ip gateway netmask From within VM i cant connect to the Internet. Do i need to change something in global zone?
[17:51:45] <serzhit> physical nic. sorry.
[17:51:45] <serzhit>  Not vnic.
[17:52:06] <serzhit> netstat -r
[17:52:28] <serzhit> hangs after table headings
[17:52:34] <wesolows> use netstat -rn
[17:52:38] <nefilim> dont resolve
[17:52:41] <nefilim> ya
[17:52:42] <wesolows> the n suppresses DNS lookups
[17:53:00] <jesse_> netstat -r should default to -n and have -N to turn on resolve
[17:53:06] *** jdefelice has quit IRC
[17:53:10] <jesse_> nobody uses netstat -r when eveything is working=)
[17:53:26] *** jdefelice has joined #smartos
[17:53:30] *** bayoda has quit IRC
[17:54:52] <wesolows> jesse_: I was just thinking the same myself.  There's never a time when I want the DNS name of a router anyway.
[17:55:00] <serzhit> netstat -rn
[17:55:01] <serzhit> Routing Table: IPv4
[17:55:01] <serzhit>   Destination           Gateway           Flags  Ref     Use     Interface
[17:55:01] <serzhit> default              UG        2         40 net0
[17:55:01] <serzhit>              UH        2          0 lo0
[17:55:01] <wesolows> and named networks are just nutty
[17:55:02] <serzhit>         U         3          1 net0
[17:55:03] <serzhit> Routing Table: IPv6
[17:55:03] <serzhit>   Destination/Mask            Gateway                   Flags Ref   Use    If
[17:55:04] <serzhit> ::1                         ::1                         UH      2       0 lo0
[17:55:09] <wesolows> pastebin!
[17:55:10] <nahamu> please use a pastesbin!
[17:55:34] *** jdefelice has quit IRC
[17:55:35] <wesolows> so, questions are the usual when debugging network problems... can you ping
[17:55:42] <wesolows> (from the zone?  from the GZ?)
[17:55:55] <serzhit> yes from zone
[17:56:06] <wesolows> then the problem is almost surely with the router
[17:56:06] *** jdefelice has joined #smartos
[17:56:19] *** bayoda has joined #smartos
[17:56:42] <serzhit> and from global zone
[17:56:57] <wesolows> if you can ping your router but nothing beyond it, it's time to look at the router
[17:56:58] <serzhit> with what router? in global zone?
[17:57:11] *** jdefelice has quit IRC
[17:57:23] <jesse_> is your global zone's address?
[17:57:23] <nahamu> the router whose IP is
[17:57:25] <wesolows> based on your configuration, I assume that is a router on that subnet with access to networks beyond
[17:57:31] *** jdefelice has joined #smartos
[17:57:39] <wesolows> if it isn't, that's a configuration error.  if it is, that router is not working properly.
[17:58:56] *** wramthun has quit IRC
[17:59:19] <serzhit> is on vnic0 in global zone
[17:59:37] <wesolows> well, that's obviously not going to work then.
[17:59:46] <wesolows> you need to use the address of a router.
[18:00:43] <wesolows> is this 192.168.1/24 network actually the subnet to which this system's physical nic is attached?
[18:00:51] <wesolows> if not, that definitely won't work.
[18:00:58] <serzhit> you mean in zone i use "gateway": real IP of physical interface here?
[18:01:01] <wesolows> you can't just make up subnets for VMs.
[18:01:14] *** jdefelice has quit IRC
[18:01:21] <wesolows> no, not the physical NIC, the router.  Some other device or machine on the network somewhere.
[18:01:27] *** wramthun has joined #smartos
[18:01:35] <wesolows> The router is not a part of the physical system running SmartOS.
[18:01:40] <Zigara> ^ now that is an interesting host
[18:01:47] *** jdefelice has joined #smartos
[18:02:35] <serzhit> Sorry I'm weak in network administration.
[18:03:02] <serzhit> I have rented dedicated server on provider premises.
[18:03:31] <wesolows> your provider should have given you a gateway or router address.
[18:03:46] <serzhit> I have one physical interface and want my zones to share the same physical interface and IP
[18:04:13] <jesse_> then you'll need to do NAT
[18:04:13] <wesolows> that's not possible, I don't believe.
[18:04:24] <wesolows> You will need to set up ipnat.
[18:04:38] <jesse_> and with it comes all the downsides of NAT
[18:04:40] <serzhit> I have gateway from provider. Do i have to put in zone's settings?
[18:04:50] <wesolows> And construct another subnet for your zones and a NATting router to allow them access to the real network.
[18:05:05] <wesolows> Yes, but it won't help if you have only a single address.
[18:05:12] <wesolows> Each zone needs its own address(es).
[18:05:33] <jesse_> serzhit, you'll want something like that: http://wiki.smartos.org/display/DOC/NAT+using+Etherstubs
[18:05:35] <rmustacc> serzhit: Yes, the zone needs to use the provider's gatway, just as the global zone does.
[18:05:36] <wesolows> Normally they would all use the same gateway.
[18:06:00] <jesse_> serzhit, (I haven't used those instructions, they may be wildly wrong)
[18:06:38] <serzhit> ))
[18:07:01] <jesse_> serzhit, easiest way is to get a subnet from your hosting provider (or bunch of IPs) and use those
[18:07:22] <jesse_> which may still not be easy and cheap=)
[18:08:48] <serzhit> yes it is much easier to get them )
[18:09:05] *** jdefelice has quit IRC
[18:09:15] <wesolows> just make sure they are all on the same subnet and you will be good to go.
[18:09:25] <serzhit> but only tomorrow ( I wanted to do some investigations this night..
[18:09:32] <wesolows> i.e., "I want N addresses on the A.B.C.D/XX subnet"
[18:09:38] *** jdefelice has joined #smartos
[18:09:38] <jesse_> wesolows, would there be problems if each vm were in its own subnet?
[18:09:42] <wesolows> whatever A.B.C.D/XX is.
[18:10:09] <wesolows> jesse_: It should still be possible to make it work, but you'd need to configure more interfaces and it would be more work.
[18:10:20] <wesolows> It's far simpler if you just have a block.
[18:10:38] <jesse_> yes, I was just wondering if there is a technical limitation
[18:10:55] *** bens1 has quit IRC
[18:11:04] <wesolows> Hell no.  Almost any network topology can work if you're willing and able to plumb up all kinds of stuff.
[18:11:32] <wesolows> But for someone who assesses his network skills as weak, the path to success runs through a single block of addresses on a single subnet.
[18:11:43] <jesse_> yes, definitely
[18:11:52] *** bens1 has joined #smartos
[18:12:32] *** bens1 has quit IRC
[18:17:17] *** wramthun has quit IRC
[18:17:25] <serzhit>  Thank you guys! I thought that there is a simple way to get running zones on single IP. There is no such way. ;)
[18:17:25] *** notmatt has joined #smartos
[18:18:41] *** jdefelice has quit IRC
[18:19:12] *** jdefelice has joined #smartos
[18:19:28] *** wramthun has joined #smartos
[18:20:05] *** twobitfool has joined #smartos
[18:20:21] *** jdefelice has quit IRC
[18:20:38] *** twobitfool has left #smartos
[18:20:54] *** jdefelice has joined #smartos
[18:21:11] *** sachinsharma has quit IRC
[18:21:31] *** two-bit-fool has joined #smartos
[18:21:38] *** miller7 has quit IRC
[18:22:52] *** jdefelice has quit IRC
[18:23:25] *** jdefelice has joined #smartos
[18:24:00] <two-bit-fool> When trying to run SmartOS in a virtual machine, after the configuration, I get the following error: mv: cannot rename /tmp_config to /usbkey/config: No such file or directly /smartdc/bin/smartos_pompt_config.sh: line 673: /usb/key/shadow: No such file or directory
[18:24:36] <two-bit-fool> And by "directly" I meant "directory" ;)
[18:26:01] <two-bit-fool> Anyone have any guesses to where I went wrong?
[18:27:29] <rmustacc> What does zpool status say?
[18:27:56] *** jdefelice has quit IRC
[18:28:00] *** jdefelice1 has joined #smartos
[18:29:54] <two-bit-fool> Well, I never get to a prompt. That error shows up right after the config process.
[18:32:14] <two-bit-fool> If I boot into single-user mode, `spool status` returns "no pools available"
[18:34:14] <e^ipi> yeah because it's not imported
[18:34:18] <nahamu> two-bit-fool: what does "zpool import" return?
[18:34:28] *** jdefelice1 has quit IRC
[18:34:53] <two-bit-fool> `zpool import` # => nothing
[18:35:25] *** jdefelice has joined #smartos
[18:39:11] *** Wraithhh has quit IRC
[18:39:59] *** jdefelice has quit IRC
[18:40:21] *** jdefelice has joined #smartos
[18:41:00] *** enmand_ has joined #smartos
[18:43:37] *** enmand has quit IRC
[18:45:38] *** Wraithhh has joined #smartos
[18:45:51] *** KermitTheFragger has quit IRC
[18:48:52] *** tonyarkles has quit IRC
[18:50:04] *** mamash has joined #smartos
[18:50:26] *** Cpt-Oblivious has joined #smartos
[18:51:07] *** Cpt-Oblivious has quit IRC
[18:51:46] *** wramthun has quit IRC
[18:51:50] *** Cpt-Oblivious has joined #smartos
[18:53:29] *** tonyarkles has joined #smartos
[18:53:42] *** sachinsharma has joined #smartos
[18:54:31] *** wramthun has joined #smartos
[18:55:41] *** jedahan has joined #smartos
[18:55:53] *** CarlosC has quit IRC
[18:57:51] <two-bit-fool> I've tried using the vmware vm (from http://wiki.smartos.org/display/DOC/Download+SmartOS), and also tried spinning up a vm on virtual box (using script from http://www.perkin.org.uk/posts/automated-virtualbox-smartos-installs.html), but in either case I get that same error (cannot rename /tmp_config to /usbkey/config).
[18:58:36] <two-bit-fool> Do I need to drop into single-user mode, and set up zfs first?
[18:58:55] <nahamu> jperkin: is the 12.4.1 multiarch to be preferred over the 12.4.0?
[18:59:04] <wesolows> more likely you need to figure out why it's not creating a zpool
[18:59:46] <jperkin> nahamu: yes
[19:00:31] <nahamu> jperkin: any objection to my adding those images to my image server?
[19:00:33] *** bens1 has joined #smartos
[19:00:43] <nahamu> (which will also trigger mirroring by datasets.at)
[19:02:11] <jperkin> nahamu: not at all.
[19:03:01] *** ipalreadytaken has joined #smartos
[19:05:59] *** enmand_ has quit IRC
[19:07:12] <nahamu> done.
[19:07:25] <chris---> two-bit-fool so you booted from an iso in a vm, and answered the questions, and it gave those errors then rebooted and doesnt take you back to the config screen?
[19:07:30] *** ipalreadytaken has quit IRC
[19:08:16] <two-bit-fool> @chris--- After the error in reboots, and then asks me the same questions all over again.
[19:08:26] *** szaydel has joined #smartos
[19:08:30] <two-bit-fool> sorry "it reboots"
[19:08:41] *** serzhit has left #smartos
[19:09:35] *** wramthun has quit IRC
[19:09:59] <chris---> the vm is brand new, the "disks" werent used in something else doing zfs before
[19:10:19] *** enmand has joined #smartos
[19:10:41] <two-bit-fool> @chris--- Yup, brand new. In the case of the VM downloaded from smartos.org, I'm just using the disks that were already defined.
[19:11:30] <jesse_> I got that error a while back, too, on real hardware
[19:11:42] *** sachinsharma has quit IRC
[19:12:35] <jesse_> I was installing on an existing zones-pool at the time (theoretically, shouldn't matter). I think it was because I already had a zones/opt fs in the pool
[19:12:36] *** wramthun has joined #smartos
[19:12:57] <jesse_> renamed the fs, removed the .dot file from zones/ and reinstall worked
[19:14:29] <jesse_> (s/existing zones-pool/existing pool renamed to 'zones' in maintenance mode/)
[19:14:38] *** tonyarkles has quit IRC
[19:14:39] *** ipalreadytaken has joined #smartos
[19:15:27] *** simong has joined #smartos
[19:16:42] <jesse_> I wonder if there's a race condition between zfs create call and it appearing in the fs and writing to said fs
[19:18:20] <chris---> the provided vmware image isnt already installed?
[19:18:23] <chris---> i havent tried
[19:18:31] <chris---> obviously
[19:19:26] <chris---> i've made a few vbox vms by just booting the iso from scratch
[19:22:27] *** CarlosC has joined #smartos
[19:22:40] <two-bit-fool> chris---: Yeah, I've tried that too, and still the same error.
[19:25:01] <chris---> you have other vms of other things that work ok
[19:25:31] <two-bit-fool> yep.
[19:25:54] <two-bit-fool> I'm just reading the source of the setup script (https://github.com/joyent/smartos-overlay/blob/master/smartdc/bin/smartos_prompt_config.sh)
[19:26:26] *** tonyarkles has joined #smartos
[19:26:27] <two-bit-fool> My error is from line 669, so the /usbkey must not be available
[19:26:48] *** simong has quit IRC
[19:27:20] <jperkin> right, somewhere along the line your disks aren't being configured correctly, /usbkey resides on them.
[19:28:17] <chris---> that area of the startup script will spit errors that are hard to see flash by
[19:30:21] <two-bit-fool> chris---: that's unfortunate
[19:30:45] <wesolows> you can start with diskinfo
[19:30:47] <wesolows> and/or disklist
[19:32:14] <two-bit-fool> something must be going bad in create_zpool, because I never see the sleep 5.
[19:34:10] *** simong has joined #smartos
[19:34:31] <rmustacc> You did see disks listed at the prompt and enteredthem right?
[19:35:12] <two-bit-fool> Yeah.
[19:38:01] <chris---> and it checks that against the found disks, so that shouldnt be able to proceed with wrong devices
[19:38:05] *** mgt576 has joined #smartos
[19:38:35] *** enmand has quit IRC
[19:39:19] <two-bit-fool> I just manually ran the code from create_zpool (in single-user mode) and it went fine.
[19:40:11] <chris---> destroy it before you reboot else it wont try to reinstall
[19:40:35] <jesse_> chris---, removing the dotfile from /zones/ is enough
[19:40:46] <jesse_> chris---, then, when it asks for disks, give it any device
[19:41:03] <chris---> i just see a zpool list looking for "zones" and a return 0
[19:41:03] <jesse_> chris---, beginning of create zpool, it checks if zones pool exists, and if it does, it's happy
[19:41:10] <chris---> if it finds it
[19:41:20] <chris---> oh
[19:41:24] <chris---> i see
[19:41:36] <jesse_> that's the way to upgrade existing zpool into a zones-pool
[19:42:19] <jesse_> boot into maintenance, zpool import yourpool zones, rename zones/opt and zones/var, reboot, give any_device as disk, done
[19:42:40] <jesse_> (it won't clobber that any_device, but it's still a bit unsettling every time to give it something=))
[19:42:42] <chris---> that was just a check to not try to make a new pool with the same name
[19:43:01] <chris---> the .system_pool check is before that script it looks like
[19:43:06] <jesse_> before imgadm import: zpool upgrade zones
[19:43:13] <jesse_> otherwise imgadm will fail=)
[19:43:16] <two-bit-fool> what's the command to restart on smartos?
[19:43:28] <jesse_> two-bit-fool, 'reboot' work for me
[19:43:55] <jesse_> chris---, yeah, you need to make sure /zones/.system_pool doesn't exist
[19:43:57] *** enmand has joined #smartos
[19:44:06] <jesse_> or you won't get to the installer
[19:44:58] <two-bit-fool> It's looking better this time...
[19:45:54] *** ipalreadytaken has quit IRC
[19:46:37] <two-bit-fool> hmm, no errors this time, but after config it reboots, forgets everything, and takes me through setup again.
[19:47:54] *** ipalreadytaken has joined #smartos
[19:52:24] *** simong has quit IRC
[20:06:07] *** bens1 has quit IRC
[20:07:18] <two-bit-fool> jesse_ & chris--- thanks for the help. i figured it out
[20:09:24] <chris---> what was it
[20:09:35] <two-bit-fool> …and boy did I earn my nickname today. I was hitting [enter] on the final confirmation, which defaults to [n].
[20:09:55] <chris---> heh
[20:10:51] <two-bit-fool> Then it gave me the "The system will now finish configuration and reboot. Please wait…" message, followed by errors, so I just assumed that something was wrong.
[20:11:29] <jperkin> well, at least you admitted it ;)
[20:12:07] <two-bit-fool> I'm looking at the script now, and wondering the best way to handle that case -- to prevent fools like me from slipping into that trap.
[20:13:00] *** abhishekvasisht has joined #smartos
[20:13:03] *** jdefelice has quit IRC
[20:13:06] <two-bit-fool> Maybe test for $val != "y" to show a "Installation cancelled" message and a quick exit stage right?
[20:13:36] *** jdefelice has joined #smartos
[20:14:09] *** abhishekvasisht has quit IRC
[20:15:34] *** notmatt has quit IRC
[20:15:45] <two-bit-fool> Or is there another case, where someone would answer [n] but still want the rest of the script to run?
[20:19:07] *** jdefelice has quit IRC
[20:19:24] <chris---> if you say no to making the data pool not sure there is any reason to continue
[20:20:04] *** jdefelice has joined #smartos
[20:21:16] *** jdefelice has quit IRC
[20:21:49] *** jdefelice has joined #smartos
[20:30:43] *** deirdres has joined #smartos
[20:38:52] *** Cpot-Oblivious has joined #smartos
[20:39:38] *** ipalreadytaken has quit IRC
[20:41:50] *** Cpt-Oblivious has quit IRC
[20:44:12] *** szaydel has quit IRC
[20:45:38] *** Cpot-Oblivious has quit IRC
[20:47:15] *** Webhostbudd has quit IRC
[20:50:02] <leecallen> I am having trouble with smbadm join to a domain.  it doesn't seem to be logging anything, or have any capability for debug or diagnostics.
[20:50:20] <leecallen> where should I look for logging?
[20:51:26] *** szaydel has joined #smartos
[20:52:00] *** iyp has quit IRC
[21:08:26] *** ipalreadytaken has joined #smartos
[21:17:10] *** miller7 has joined #smartos
[21:17:24] *** mgt576 has quit IRC
[21:17:31] *** miller71 has joined #smartos
[21:18:25] *** ipalreadytaken has quit IRC
[21:21:34] *** miller7 has quit IRC
[21:25:20] *** enmand has quit IRC
[21:26:09] *** notmatt has joined #smartos
[21:30:48] *** notmatt has quit IRC
[21:32:56] <two-bit-fool> jesse_ & chris--- Thanks again. I've submitted a pull request, which should help keep any fools from following me over that little cliff. (https://github.com/joyent/smartos-overlay/pull/18)
[21:41:07] *** mamash has left #smartos
[21:50:32] *** mamash has joined #smartos
[21:52:37] *** neophenix has quit IRC
[21:56:30] <wesolows> I will take a look at the pending overlay stuff probably early next week.  Thanks for working on this.
[21:57:15] *** two-bit-fool has left #smartos
[22:01:47] *** d[^_^]b has quit IRC
[22:01:55] *** d[^_^]b has joined #smartos
[22:03:21] *** orangeroo has joined #smartos
[22:04:40] *** ryancnelson has joined #smartos
[22:07:15] *** mamash has left #smartos
[22:08:34] *** mamash has joined #smartos
[22:12:55] *** mamash has left #smartos
[22:16:18] *** ryancnelson has quit IRC
[22:22:40] <jesse_> anybody with access to many opensolaris derivatives? what do their "head -1 /etc/release" say?
[22:23:09] <jperkin>   OmniOS v11 r151002
[22:23:23] <jesse_> (believe it or not, that's what's used to detect solaris/opensolaris in jre font selection...)
[22:24:19] *** CarlosC has quit IRC
[22:24:22] <jesse_> so, Illumos/OpenSolaris/OmniOS/SmartOS/OpenIndiana is probably a good start in detecting them
[22:24:59] <jperkin> why does it need to be different between solaris and others?
[22:25:28] <jperkin> i.e. why not just uname -s == SunOS and be done with it?
[22:26:03] <jesse_> because of different fonts
[22:26:27] <jesse_> it needs to map java-names of fonts to real fonts
[22:27:06] *** autif has quit IRC
[22:27:28] <jesse_> actually, I'll need to check what that "isOpenSolaris" does
[22:29:28] <wesolows> SmartOS x86_64
[22:29:41] <jesse_> the whole font handling is a mess
[22:29:50] <jesse_> but that's status quo for anything font-related
[22:30:12] <wesolows> one would hope that could be done by looking at, you know, the fonts available on the system
[22:30:19] <wesolows> I know, I know.
[22:30:39] <jesse_> and how many ways are there to do that on X11?=)
[22:31:17] <wesolows> presumably one could just ask the X server for a font and see if it succeeds?
[22:31:30] * wesolows is probably being very naive here
[22:33:39] <jesse_> obviously you have never worked with fonts on X11
[22:34:33] <jamesd_> the client and server may not be on the same host or platform... i run X remotely a lot for work things
[22:34:35] <Triskelios> uh, fontconfig is basically standard now
[22:34:45] <Triskelios> server-side fonts are dead
[22:35:01] <wesolows> jesse_: no, I haven't, except as an end user.
[22:35:52] <jesse_> http://hg.openjdk.java.net/jdk7u/jdk7u/jdk/file/tip/src/solaris/classes/sun/awt/fontconfigs/solaris.fontconfig.properties and http://hg.openjdk.java.net/jdk7u/jdk7u/jdk/file/tip/src/solaris/classes/sun/awt/fontconfigs/linux.fontconfig.Ubuntu.properties give a little hint what it's with java
[22:36:42] <jesse_> Triskelios, I just loved sun for having the licensed fonts as java fonts on solaris.
[22:37:11] <jesse_> Triskelios, xfs would only serve them to localhost, totally negating the benefit of an xfs
[22:37:45] <jesse_> and making it impossible to run java GUI software from solaris box to any other unix
[22:37:50] <jesse_> +screen
[22:37:52] <Triskelios> jesse_: the licenced fonts which made Java apps look different from everything else... yeah, good job
[22:37:52] <wesolows> probably because if you exposed xfs, it would immediately be used to compromise the system.
[22:39:20] <jesse_> I wouldn't say java GUI was entirely bad idea, but sun sure made all the effort it could to kill it=)
[22:39:48] <jesse_> (talking about AWT days more here)
[22:40:35] <orangeroo> Hi all
[22:42:01] <orangeroo> So I'm installing SmartOS, but after reboot it goes back into the install again.
[22:42:26] <orangeroo> Like it can't find the zones pool it made.
[22:42:43] <rmustacc> If you boot noimport and run zfs import do you see anything?
[22:43:02] <orangeroo> lemme try
[22:45:10] <jesse_> just that you're not repeating the problem before, did you answer yes to the final confirmation and not enter, as it defaults to [n]?=)
[22:45:40] <orangeroo> Oh yeah.
[22:47:35] <orangeroo> Sometimes it works, sometimes it doesn't.  That is driving me crazy.
[22:48:48] <rmustacc> Well, then it most likely is the following bug which just got fixed in illumos:
[22:49:15] <orangeroo> :\
[22:49:22] <rmustacc> https://www.illumos.org/issues/3422
[22:49:34] <rmustacc> It was fixed yesterday.
[22:49:44] <rmustacc> But wesolows would have more state on that.
[22:49:46] <orangeroo> Would I be better off using an older SmartOS version?
[22:50:01] <rmustacc> That though would only be the case if you're creating and you find that you have a created zpool that was corrupt, that's all.
[22:51:36] <wesolows> the only workarounds are in SDC
[22:51:42] <wesolows> afaik we never implemented one for SmartOS
[22:52:07] <wesolows> you would have to use something VERY old, as this bug's been around for a while.
[22:52:21] <rmustacc> Oh well, it's fixed, we should get it pulled in with the sync.
[22:53:35] <orangeroo> Oh snap.  It says my zil is corrupt
[22:54:28] <orangeroo> Doing ZIL mirror on SSD
[22:54:48] <wesolows> yeah that's 3422
[22:54:56] <wesolows> there's nothing wrong with your HW
[22:55:03] <wesolows> or the data, for that matter.
[22:56:33] <orangeroo> http://gifsoup.com/view2/1667406/headdesk-o.gif
[22:57:31] <jesse_> wesolows, so, basically, 'set a pool property'
[22:58:18] <wesolows> correct
[23:00:22] <jperkin> nahamu: multiarch packages for everything except libspice are up, am just fixing up the libspice failure (additional PKG_DEVELOPER checks)
[23:02:42] *** enmand has joined #smartos
[23:10:06] *** simong has joined #smartos
[23:18:12] *** orangeroo has left #smartos
[23:18:28] *** tonyarkles has quit IRC
[23:25:41] *** jdefelice has quit IRC
[23:26:18] *** jdefelice has joined #smartos
[23:26:39] *** tonyarkles has joined #smartos
[23:29:42] *** jdefelice has quit IRC
[23:29:56] *** jdefelice has joined #smartos
[23:35:30] *** jdefelice has quit IRC
[23:35:59] *** jdefelice has joined #smartos
[23:41:48] *** nefilim has quit IRC
[23:45:46] *** szaydel has quit IRC
[23:46:52] *** szaydel has joined #smartos
[23:50:28] *** jdefelice has quit IRC
[23:57:24] *** nefilim has joined #smartos

   February 1, 2013  
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | >