July 9, 2011 
< | 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 | 29 | 30 | 31 | >
July 9, 2011 [00:00:35] <RoyK> DontKnwMuch: I'd say use 4GB or so for SLOG unless the system has a lot of writes - use the rest for L2ARC[00:00:38] <DontKnwMuch> I did a scrub with 500GB of files on it, 450mb/s it was... perhaps it was ok speed as it was still mostly empty[00:01:35] <RoyK> DontKnwMuch: L2ARC won't help with scrub speed, but it'll help other system performance during scrub/resilver[00:01:41] <johnny_N30> my spare parts: http://pastie.org/2185129[00:02:18] <RoyK> johnny_N30: looks good[00:02:28] <DontKnwMuch> will try it. It is a shame to use only 8gb out of 40gb...[00:02:51] <RoyK> DontKnwMuch: my point exactly :)[00:02:57] <DontKnwMuch> what about this trim thing for ssd drives, they will get slower and slower and slower, or I am off[00:03:34] <RoyK> DontKnwMuch: trim isn't that important when using it for caching - it's worse if using it for filesystems[00:04:01] <johnny_N30> thanks, more details http://pastie.org/2185132. Should I be worried about Illegal Requests for USB drives?[00:04:43] <RoyK> probably not - no hard errors as far as I can see[00:05:20] <DontKnwMuch> strangely I also have illegal request .. well at least while still on ich10.. didnt check now... what does it actually mean... strange thing[00:10:11] <RoyK> DontKnwMuch: did you try to slice up those SSDs?[00:10:46] <DontKnwMuch> nope. But will try.[00:11:30] <RoyK> just zpool remove that mirror of slogs[00:11:58] <RoyK> run format/partition and create a small, equally sized slice on each drive and a bigger for the rest[00:15:00] <DontKnwMuch> thinking if I should stop the file copy running now, or test what happens ;)[00:15:27] <RoyK> removing the SLOG won't hurt the system[00:16:35] <DontKnwMuch> ok, I did it, and files are still copying.. ;)[00:17:39] <DanaG> One thing that'd be good as cache, perhaps: the Intel SSD 311.[00:17:46] <DanaG> It's SLC, so it shouldn't wear as quickly.[00:18:26] <DontKnwMuch> how fast do these mlc drive wear out actually?[00:18:28] <DanaG> Soooooooooooooooooo, where do I get gcc 4.3?[00:18:31] <RoyK> DanaG: SLC is good for SLOGs[00:18:40] <RoyK> DanaG: for L2ARC, anything should work[00:18:55] <RoyK> DanaG: if they wear out too quickly, there's always a warranty...[00:19:22] <RoyK> DanaG: and if you get bit errors on L2ARC, ZFS will just read from the drives[00:19:24] <DanaG> Intel warranty states: 3 years, or when wear counter goes to zero. Whichever is sooner.[00:19:47] <DanaG> So anyway, I needeth a GCC.[00:19:52] <DanaG> Eth.[00:20:07] <richlowe> gcc4.3 should be in the 'opensolaris' repo[00:20:12] <RoyK> DanaG: do you have a link on that?[00:20:14] <richlowe> but it's also, probably, GCCFSS[00:20:49] <DanaG> Nope, 4.3 seems to be MIA.[00:21:15] <DanaG> er, wait, maybe it's back now.[00:21:45] <DanaG> sudo pkg install gcc-43[00:21:45] <DanaG> No updates necessary for this image.[00:21:49] <DanaG> same for 432[00:22:51] <RoyK> DanaG: I just checked and can't find anything in the warranty docs about wear counters[00:22:57] <richlowe> DanaG: you know it's hidden off in /usr/gcc right?[00:23:07] <richlowe> "No updates available" suggests you already installed it[00:24:04] <RoyK> imho gcc should be in path[00:24:15] <RoyK> either installed under /usr/bin or at least symlinked there[00:25:34] <dkeav> err it is[00:25:47] <dkeav> but its sol gcc3[00:29:25] *** merzo has quit IRC[00:30:18] <DontKnwMuch> I know dick about partitions obviously, finally managed to get two on each ssd[00:30:39] <DontKnwMuch> now how can I add them as slog/l2arc[00:31:07] <DontKnwMuch> how to 'address' them[00:31:36] <Wraithh> c1t2d0s3 or so[00:32:01] *** johnny_N30 has quit IRC[00:32:40] <DontKnwMuch> they have such beautifull short names: c1t5000CCA225C0A327d0s1 ... argh[00:32:50] <Wraithh> hehe[00:32:56] <dkeav> are you going to mirror?[00:33:31] <dkeav> i also take it this is not a production server?[00:33:41] <DontKnwMuch> I had both in mirror for slog[00:33:44] <Wraithh> I wrote a script to resolve the physical sas topology and enclosure/bay mappings for the device names. Then it makes a sed rule file I can use to pipe stuff through[00:33:52] <DanaG> RoyK: maybe that was some other manufacturer, then.[00:34:03] <DontKnwMuch> no, it is not, yet[00:35:07] <DontKnwMuch> I just use napp-it, and click with the mouse ;)[00:35:11] *** raichoo has quit IRC[00:37:15] <DanaG> What do I do to add it to the path? Do I need the libs, too?[00:39:35] <RoyK> DanaG: export PATH=$PATH:/usr/gcc/4.3/bin[00:41:35] <DontKnwMuch> slices start with 0 or 1? hm... 0 I think[00:41:52] <RoyK> yes[00:43:05] *** ivo_ has quit IRC[00:46:11] <DanaG> what's the default port for napp-it?[00:47:07] <dkeav> 81[00:47:08] <DontKnwMuch> 81[00:47:25] <DanaG> okay, so libtorrent needs libboost... that's not in the repos, either.[00:50:28] <herzen> DanaG: Boost is here: http://staticdev.uk.openindiana.org:10003[00:50:40] <DontKnwMuch> I now have a 8gb slice from both drives mirror as log added... l2arc does not need to be a mirror at all, can I just use one disk slice 1 for it[00:50:44] <tomww> there is the experimental repository with boost with gcc as SFEboost-gpp and the other veriant is SFEbost-stdcxx[00:50:45] <herzen> but only built with runtime/gcc, from the same repo[00:50:54] <tomww> hey herzen :)[00:51:35] <herzen> tomww: hey[00:51:59] <tomww> I was making marketing for the repo right now[00:52:13] <herzen> lol[00:53:10] <RoyK> DontKnwMuch: 8GB is overkill unless you need very high write speed. For l2arc, use both slices and they will be striped[00:53:11] <DontKnwMuch> what does this mean (cache) that I have a raid0 sortof? http://pastie.org/2185309[00:53:29] <RoyK> looks good[00:54:03] <DanaG> herzen: if I use that, I'll have to use the whole system from there.[00:54:11] <RoyK> and yes, the l2arcs (aka cache) will be striped (aka raid0)[00:54:36] <DontKnwMuch> At least I learned something new :) now will see how long these drives will last[00:55:40] <DontKnwMuch> if one failes, I will have to add a new one, partition it and.. hm... or just remove cache/log, and add it back as I did now[00:55:42] <RoyK> DontKnwMuch: I have a very busy server at work with some 10 256GB C300 SSDs for L2ARC, one died after a few days, the rest have lived on for half a year or so, only one showing some errors[00:56:10] <herzen> DanaG: well, there's a spec at SFE that builds Boost with stlport4. that should be compatible with a run of the mill Solaris environment.[00:56:22] <RoyK> DontKnwMuch: since you're using slices, you'll need to reattach the other half of the slog[00:56:42] <RoyK> a mirror must be handed manually[00:56:44] <DanaG> I don't have a specific need to stick with the old, as long as the new is also usable. I'm not familiar with the development workflow here.[00:56:52] <RoyK> for the slog, just remove it and reattach something else[00:56:55] <herzen> but it's easier to download binary packages than it is to build packages from specs yourself[00:57:26] <DanaG> So should I just upgrade the system from that repo?[00:58:12] <DanaG> Or is that asking for trouble?[00:58:21] *** hsp has quit IRC[00:58:27] <DontKnwMuch> royk: ok... or I can remove the slog completely, and add it as I did now with the new drive partitioned[00:58:37] <tomww> this SFE stuff is additional, so the term would be instal those packages from that repository, you first have to add it as another publisher[00:58:50] <DanaG> I already added it, but have it disabled for now.[00:58:55] <tomww> so, ad additinal repo, it is not asking for trouble .)[00:59:31] <DanaG> Is it an equivalent of the "proposed" repository in Ubuntu?[00:59:37] <tomww> enable or disabled should have not special effect, since it would ony get involved if you want a package only found there[00:59:40] <DanaG> Or is it more like the +1?[01:00:20] <tomww> it is totally "on top" of what OI provides, normally nothing which competes with the core OI distribution[01:00:35] * herzen doesn't know Ubuntu[01:02:34] <DanaG> wtf? When I searched for something, it appeared during search, then disappeared after the search finished![01:03:45] <herzen> DanaG: also, are you trying to build libtorrent or libtorrent-rasterbar?[01:03:56] <DanaG> Libtorrent-rasterbar, included in Deluge.[01:04:35] <herzen> I believe i got that to build against SFEboost-stdcxx, but then dropped that, since I couldn't get qbittorrent to build[01:04:47] <DanaG> I'll be using remote UI.[01:05:16] <DanaG> Argh, I could've sworn there was a "boost" in the extra repo... now it's not there.[01:05:25] <DanaG> or it's there during search, but goes away when search "finishes"[01:06:39] <herzen> there are two "extras" repos: the other one is http://staticdev.uk.openindiana.org:10002. that has SFEboost-stdcxx[01:07:04] <DanaG> Can you explain the strange search behavior?[01:07:46] <herzen> no. the packages in the two repos are named differently, BTW. that's the main difference between them.[01:07:51] <DanaG> Oh yeah, so my question of analogies: +1 is the next major version (6 month release); "proposed" is small upgrades that need testing before being pushed out to everyone.[01:08:16] <DanaG> Or is this just another, approximately-same version, third-party repo?[01:08:30] <herzen> no, most of it is something that probably will never get into OI[01:09:23] <herzen> these are the flags I used to build libtorrent-rasterbar:[01:09:38] <herzen> export CXXFLAGS="-xtarget=native -xO3 -library=stdcxx4"[01:09:38] <herzen> export LDFLAGS="-library=stdcxx4 -L/usr/stdcxx/lib -R/usr/stdcxx/lib"[01:09:38] <herzen> ./configure --with-boost=/usr/stdcxx --disable-static --enable-python-binding --with-libgeoip[01:11:35] *** tsoome1 has joined #openindiana[01:12:09] *** tsoome has quit IRC[01:12:21] *** tsoome has joined #openindiana[01:16:11] *** tsoome1 has quit IRC[01:16:53] *** synegy34 has joined #openindiana[01:21:20] <DanaG> hmm, there's no .configure in the libtorrent dir in Deluge.[01:22:42] <RoyK> DontKnwMuch: yes[01:23:56] <sergefonville> how do increase the max open files?[01:24:06] <DanaG> autoheader: error: AC_CONFIG_HEADERS not found in configure.in[01:24:42] <DanaG> ah, I just had to run autoconf.[01:26:47] <DanaG> Argh, nope, still no dice.[01:26:54] <DanaG> All sorts of missing commands.[01:28:04] <DanaG> ./configure: line 1798: syntax error at line 1801: `(' unexpected[01:31:59] <DanaG> automake: command not found[01:33:59] <DanaG> Okay, so now I'm totally boggled.[01:38:57] <DanaG> s/boggled/lost/[01:42:22] <DanaG> What's the difference between SFEBoost-stdcxx-devel and system/library/g++/boost?[01:43:08] <sergefonville> alternatively, how do I compile nginx 64-bit?[01:50:11] <DanaG> ah, just had to edit setup.py to add -I/usr/stdcxx/include[01:52:47] <DanaG> /usr/gcc/4.3/lib/gcc/i386-pc-solaris2.11/4.3.3/../../../../include/c++/4.3.3/bits/boost_concept_check.h:216: error: ‘__gnu_cxx::_SGIAssignableConcept<_Tp>::__a’ has incomplete type[02:01:18] <DanaG> Argh, this isn't working![02:03:14] *** hajma_ has quit IRC[02:06:30] *** hajma_ has joined #openindiana[02:07:38] <DanaG> argh, got it to work.[02:07:44] <DanaG> had to edit setup.py to add includes.[02:07:53] <DanaG> Also had to use 4.6, not 4.3.[02:12:30] <DanaG> Okay, it built fine... but now ImportError: ld.so.1: isapython2.6: fatal: libboost_filesystem.so.1.46.1: open failed: No such file or directory[02:12:38] <DanaG> and I need to get ready to go soon/now.[02:16:18] <DanaG> bah, I'll just take the easy way and symlink those.[02:17:20] <DanaG> ImportError: ld.so.1: isapython2.6: fatal: relocation error: file /usr/lib/libboost_filesystem.so.1.46.1: symbol __1cFboostGsystemPsystem_category6F_rkn0BOerror_category__: referenced symbol not found[02:17:23] <DanaG> AAAAAAAAAAAAAAAAAARRRRGH[02:17:33] <DanaG> Anyway, enough of this for now.[02:17:42] *** DanaG has quit IRC[02:27:26] *** redgone has joined #openindiana[02:33:54] *** InTheWings has quit IRC[02:40:05] *** McBofh has quit IRC[02:49:11] *** Botanic_ has joined #openindiana[02:49:42] *** Botanic has quit IRC[02:50:51] *** redgone has quit IRC[02:56:59] *** sergefonville has left #openindiana[03:10:47] *** master_of_master has quit IRC[03:12:58] *** master_of_master has joined #openindiana[03:18:03] *** gea__ has quit IRC[03:19:45] *** ThomasB2k has joined #openindiana[03:36:47] *** Norrin has quit IRC[03:37:03] *** Norrin has joined #openindiana[04:49:43] *** McBofh has joined #openindiana[05:07:49] *** descipher has quit IRC[05:07:51] *** descipher_ has joined #openindiana[05:07:54] *** descipher_ is now known as descipher[05:12:29] *** DrLou has quit IRC[05:22:12] *** mnaser_ has joined #openindiana[05:22:13] *** mnaser has quit IRC[05:22:13] *** mnaser_ is now known as mnaser[05:31:13] *** mnaser has quit IRC[05:31:51] *** mnaser has joined #openindiana[05:53:40] *** dijenerate has joined #openindiana[06:02:46] *** DanaG has joined #openindiana[06:03:59] <DanaG> Oookay, so after all my wrangling to get Deluge to compile, the daemon won't start.[06:04:04] <DanaG> ImportError: ld.so.1: isapython2.6: fatal: relocation error: file /usr/lib/libboost_filesystem.so.1.46.1: symbol __1cFboostGsystemPsystem_category6F_rkn0BOerror_category__: referenced symbol not found[06:05:38] <DanaG> Say, if I make an old-version pool on a partition, can I later install into that root pool?[06:14:42] *** kdavy has joined #openindiana[06:15:58] *** tsukasa has quit IRC[06:19:39] <DanaG> So, if I can get that, and some DAAP share app, then it's a winner.[06:22:57] <DanaG> oh heck, there's a Transmission client.[06:22:59] <DanaG> But no daemon.[06:24:25] <Triskelios> transmission-daemon, as always[06:24:36] <Triskelios> were you trying to update SFEdeluge? that was built with Sun Studio[06:24:38] <DanaG> weird... alt-f2 didn't show it.[06:24:47] <DanaG> Triskelios: nope, I was trying to build from source.[06:25:07] <Triskelios> please look at SFEdeluge[06:26:25] <DanaG> hmm, no such package in sfe or oi-sfe.[06:26:31] <DanaG> er, sfebuild or oi-sfe.[06:27:22] *** TPickle has quit IRC[06:27:29] <DanaG> I'll also check Transmission.[06:27:48] <DanaG> My problem with Transmission on FreeNAS is that it kept insisting some files were 0% downloaded.[06:28:14] <Triskelios> FreeNAS isn't illumos/Solaris based[06:28:32] <DanaG> So maybe I'll try it on OI.[06:29:02] <DanaG> How about DAAP shares?[06:29:18] <DanaG> I do like the idea of a complete fault management framework.[06:29:50] <Triskelios> I think rythmbox has some DAAP stuff, not sure[06:29:55] <DanaG> OI also seems to have rather sparse repos -- for example, no netatalk.[06:30:03] *** tsukasa has joined #openindiana[06:30:16] <Triskelios> DanaG: look at SFE itself, not the binaries[06:30:54] <DanaG> http://pkgbuild.sourceforge.net/spec-files-extra/[06:31:01] <Triskelios> you're one of those people using netatalk for Time Machine or something? :P[06:31:43] <DanaG> Considering it. Not certain yet.[06:32:04] <DanaG> I'm actually thinking I'll instead have my nas's copy of the stuff be its own thing, and leave the direct-attached time machine drive.[06:34:07] <DanaG> Oh, and one gripe: how do I get home and end keys to not just insert ~~~~~~~~~~~~ ?[06:37:37] <DanaG> Another thing was that my chosen USB stick, that works fine with Linux, is slow as molasses with OI (as boot drive).[06:37:48] <DanaG> Hah:[06:37:50] <DanaG> Error:[06:37:50] <DanaG> This is an Live Image. The install operation can't be performed.[06:37:57] <DanaG> That's on the install within my vbox.[06:38:48] <alanc> yes that's a known bug[06:38:58] <alanc> the solution is to use CLI instead of GUI[06:39:17] <DanaG> Last time I tried CLI install, it didn't create my user account.[06:39:25] <alanc> or poke OI to incorporate the fix for that bug that upstream made months ago[06:39:51] <alanc> not CLI install of the OS, the pkg command line instead of the Package Manager gui to install additional packages[06:40:13] <DanaG> ah.[06:40:18] <alanc> and if it's gnome-terminal where home & end are failing, check the preferences - I think it's one of the compatibility setting[06:42:58] <DanaG> ah, or maybe pgup and pgdn were what I meant.[06:44:22] <alanc> yeah, there used to be a setting to choose between pgup & pgdn scrolling or being sent to the terminal for apps to handle[06:44:41] <alanc> if they're going to the terminal, shift-pg-up and shift-pg-dn should scroll[06:45:22] <alanc> though I don't see those in current gnome-terminal, but gnome likes hiding options[06:47:24] <DanaG> I mean to move on same line.[06:48:12] <alanc> that would be home & end then, though I've always used the emacs bindings instead (Ctrl-E and Ctrl-A) out of long habit[06:48:32] * DanaG forgot what pgup/pgdn even do on Linux in gnome-terminal.[06:48:35] <alanc> home & end seem to work fine for me though[06:48:37] <DanaG> I think they beep, or something.[06:49:02] <DanaG> What I was getting that was more annoying, was backspace inserting ~ and delete backspacing.[06:49:42] <alanc> Edit -> Profile Preferences -> Compatibility -> Backspace key / Delete key generates[06:51:10] <DanaG> yup, did that the first time you suggested it -- thanks.[06:51:31] *** jpg has quit IRC[06:53:24] *** Naresh has joined #openindiana[07:01:01] <DanaG> The action 'dir group=other mode=0755 owner=viskov path=opt/dtbld/lib/cbe-env' in package 'pkg://oi-sfe/cbe/desktop/CBEenv at 1 dot 8.0,5.11-0.147:20110410T151801Z' has invalid attribute(s): '//opt/dtbld/lib/cbe-env' cannot be installed; 'viskov' is an unknown or invalid user.[07:01:38] <DanaG> lol. oh wait, I just should download pkgtool directly.[07:02:48] <DanaG> http://pkgbuild.sourceforge.net/spec-files-extra/[07:03:05] <DanaG> Not sure what else to do... all the other links are redirected, thanks to oracle.[07:03:10] <alanc> yeah, sounds like someone screwed up building that package for the oi-sfe repo[07:04:57] *** Avantopia has joined #openindiana[07:06:20] <DanaG> Do I really need that "sun studio" that's no longer findable as "sun studio"... or is GCC 4.6 okay?[07:06:33] <DanaG> er, wait, there's a sunstudio package.[07:06:35] <DanaG> Silly me.[07:06:41] <alanc> depends what you're trying to do[07:06:52] <alanc> lots of code works with either, some doesn't[07:07:32] <DanaG> I'm trying to build the spec for deluge.[07:07:33] <alanc> the kernel for example, doesn't yet build with gcc4[07:07:42] <DanaG> And try to find some DAAP streaming thingy.[07:08:27] <DanaG> oh, and something's eating 100% guest CPU, yet doesn't show in gnome-system-monitor.[07:09:13] <DanaG> oh, it was metacity.[07:09:23] <DanaG> When compiz is running, metacity eats cpu.[07:09:49] <alanc> metacity should be killed when you switch to compiz[07:11:11] <DanaG> It wasn't, for some reason.[07:11:51] <DanaG> Oh, and the USB slowness would be nice to fix, as well.[07:12:19] <DanaG> My 8GB usb drive is slower with OpenSolaris than same drive with any other OS, or other drive with OpenSolaris.[07:12:24] <DanaG> s/solaris/indiana/[07:12:30] <DanaG> Who decided on that state name, anyway?[07:13:38] <DanaG> oh, and I'd love to have a tool like "htop" (top doesn't cut it), and iotop (iostat is not the same), and iftop.[07:14:21] <DanaG> oh, iftop is there![07:18:56] <alanc> Ian, who lived in Indiana, and liked starting distros with his name hidden in their name[07:19:29] <alanc> well, if you can consider two a pattern[07:20:39] <alanc> it was just a code name, that faded away, then got a second life when the OI project needed to pick a name for their fork[07:21:21] <alanc> dunno what htop does, but there's also prstat for process monitoring[07:23:22] <DanaG> htop is like an extra-fancy "top".[07:23:31] <DanaG> Colors, nice columns, and bars for memory usage and such.[07:26:52] <alanc> ah, dunno anything like that then[07:28:06] <DanaG> sfedeluge is 2 years old.[07:28:33] <DanaG> I think I'll try Transmission first. Or build deluge with sunstudio.[07:31:09] <DanaG> Now, I'm not sure where I'd set transmission to run as a system daemon.[07:31:48] <DanaG> Or rather, what user would I run it as?[07:32:01] <DanaG> oh, these are questions for #transmission[07:33:07] <DanaG> Okay, so there's just the USB horrible slowness left to deal with.[07:34:37] *** mnaser has quit IRC[07:34:44] <DanaG> ah: https://trac.transmissionbt.com/wiki/HeadlessUsage[07:34:47] <DanaG> So yeah, just USB left.[07:35:30] *** yalu has quit IRC[07:37:35] *** yalu has joined #openindiana[07:43:55] <DanaG> well, that spec seems broken.[07:44:22] <DanaG> basename: missing operand Try `basename --help' for more information. and sh: line 1: g++: not found[07:48:48] <DanaG> pkgbuild: checking for C compiler default output file name... configure: error: C compiler cannot create executables[07:49:58] <DanaG> I used arrow up, then home to reach beginning of line, and typed, and got this garbage:[07:49:58] <DanaG> todepsicroserver:~/Downloads/SFE$ ppkgtool bui pkgtool build --download SFEmt-daapd.spec --autodepswnload SFEmt-daapd.spec --autodepswnload SFEmt-daapd.spec --autodeps[07:52:30] <DanaG> My lines of text are wrapping back to the beginning of the same line, even when not even reaching the width of the window![07:55:38] <DanaG> oh, and how do I get a VT?[07:57:07] <alanc> you have to enable the VT services[07:57:22] <alanc> http://hub.opensolaris.org/bin/view/Project+vconsole/[07:57:27] <alanc> has instructions[08:00:13] <DanaG> Thanks.[08:01:38] *** johnny_ has joined #openindiana[08:03:28] *** johnny_ is now known as johnny_N30[08:05:51] <DanaG> ooooooookay, so anyway, that manifest is broken, it seems. Or something is broken.[08:07:03] <DanaG> argh, consoles are annoying^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H^H[08:07:10] <DanaG> Because you get that, instead of actually backspacing.[08:07:36] <DanaG> At the username prompt, that is.[08:09:50] <alanc> stty erase[08:10:04] <alanc> well once you login[08:10:12] <alanc> I forget which /etc file to edit to change the default[08:12:00] <DanaG> Tons of basename: missing operand.[08:12:06] <DanaG> Any better way to install mt-daapd?[08:12:59] <DanaG> oh, it's looking for a CC variable.[08:13:08] <alanc> sorry, never heard of it[08:13:31] <DanaG> oh, and home/end on tty insert escape character garbage.[08:14:26] <alanc> which is probably part of the reason I never got in the habit of using them, as they're so often useless[08:15:36] <DanaG> I've gotten used to using them extensively, on Linux.[08:17:19] <DanaG> oh, and shift+pgup/pgdn doesn't work in tty.[08:17:29] <DanaG> I took "them" to mean keys, not TTYs.[08:17:38] <DanaG> But same is true of both in Linux -- they're useful.[08:18:24] *** jpg has joined #openindiana[08:20:19] <DanaG> And since pkgtool is perl, I can't bash -x it.[08:22:24] <DanaG> oh, perl -d[08:22:25] <DanaG> then 's'[08:25:19] <DanaG> nope, it fails to even run when called that way.[08:25:53] <alanc> oh, nothing will cause the solaris kernel console to scrollback, it's simply got no scrollback[08:25:58] <alanc> use X if you want to scroll[08:25:59] <DanaG> so how do I debug that spec file?[08:26:28] <DanaG> It's obviously broken -- it's not even setting $CC, somehow[08:26:55] <alanc> CC should be set by the env script you run to set up your environment for pkgbuild[08:27:01] <alanc> it assumes you've done that[08:27:26] <alanc> if you installed the CBE package, it would be /opt/dtbld/bin/env.sh[08:28:51] <DanaG> Before this, the most I'd done with Solaris was to compile my own programs, for a class, with my own makefiles.[08:29:14] <DanaG> Oh yeah, cbe is what wouldn't install.[08:29:53] <DanaG> What's the meaning of this section in the spec file, then?[08:29:55] <DanaG> %build export CC=gcc[08:30:07] <DanaG> replaced linebreak with lots of spaces, by the way.[08:30:27] <alanc> beats me, I rarely use specfiles, don't know what most stuff in them does[08:33:42] <DanaG> argh, that's one thing that's blocking me from going with OpenSolaris: it's a pain to build stuff for it.[08:34:03] <miine> DanaG: wait until you want to package your builds ;-)[08:34:06] <DanaG> Whereas with FreeNAS, most of what I want is already there, aside from the Time-Slider thing and fmd.[08:34:14] *** DeanoC has joined #openindiana[08:34:14] *** ChanServ sets mode: +o DeanoC[08:35:19] <miine> remember me warning you of those "surprises" ? building stuff is one of them...[08:35:34] <miine> but after some time you get used to it.[08:35:48] <miine> first of all - if you don't know spec files yet - forget them.[08:36:13] <DanaG> I don't intend to package anything... I just want to install it.[08:36:28] <DanaG> And that's part of what'd make a custom-live even more of a hassle.[08:36:49] <miine> just look what compiler settings they use. maybe you can find some also at opencsw.org etc.[08:37:14] <miine> DanaG: do you build within a zone?[08:37:43] <DanaG> I have no clue -- I don't even know what a zone is.[08:38:11] <DanaG> And this page, with direct links to february 2010... aren't there newer versions?[08:38:11] <DanaG> http://wiki.openindiana.org/oi/Building+SFE+Packages[08:38:16] <DanaG> Of CBE.[08:38:19] <miine> DanaG: so take my advise: make yourself familar with solaris zones. and setup your build environment within a zone.[08:38:27] <miine> do this now.[08:40:30] <miine> just trust me - everything else is not that important now, because you need to learn how to use zones where you can fuck up things at your will wihout affecting everything else... (whats exactly needed for building solaris stuff...)[08:43:58] <DanaG> Frankly, it's looking like really high cost, for unknown rewards for my purposes.[08:45:01] <miine> DanaG: your rewards are the best possible technology out there: ZFS, zones, dtrace, crossbow etc. . but its a steap learning curve at the beginning.[08:45:40] <DanaG> Hmm, so maybe I'll have my NAS stick with FreeNAS, and do any Solaris experimentation in VirtualBox.[08:47:00] <bdha> Read the Sysadmin Guides.[08:47:07] <miine> if it takes the pressure of having it "running now", do it. or run that apps that you can't get running nativly in solaris within virtualbox (export the data via nfs into it...)[08:47:54] <DanaG> It'll also be storage for my parents' stuff, so I'll want to avoid experimenting on that at all.[08:48:42] <miine> maybe we will get back the linux zone brand later: with that you could create a linux zone inside opensolaris. and it worked really well.[08:49:20] <miine> DanaG: the storage side (filesharing) is save on oi. just export via cifs or nfs. its a no brainer...[08:50:30] <DanaG> hmm, maybe minimal solaris, plus vbox for the apps, is okay after all.[08:50:42] *** freedomrun has joined #openindiana[08:51:03] <DanaG> But, I'll want to either install to USB, or have rpool be mirrored. The latter sounds more useful -- back up everything all at once.[08:51:21] <DanaG> And then I won't have to build anything.[08:51:36] <DanaG> Is the "time slider" thing exposed in any way, such as special samba folders?[08:52:18] <DanaG> oh, and I'll need to make sure the ZFS pools are 4K-aligned. The Samsung drives lie about block size -- both physical and virtual claim to be 512.[08:52:20] <miine> mirrored rpool is fine, but not neccessary. you can do a zfs send -R rpool > xxx.txt and have this way a backup of your rpool on the data pool...[08:52:45] <miine> forget that 4K thing. you may loose 10% write performance - so what?[08:52:56] <DanaG> I remembered hearing about it being more like 30% in some cases.[08:53:22] <DanaG> The mirror thing was more about this: I don't want to have an extra spinning drive for boot.[08:53:25] <miine> I don't think that were studies using even near scientific methods...[08:53:42] <DanaG> I'd like to use USB, but my only big USB stick is slow in Solaris. It's fine in Linux, though![08:54:04] <DanaG> Or seems to be, at least. Slow, but not as bad as in Solaris, where it's like molasses.[08:55:13] <miine> put in something cheap like a used 2.5 hd . rpool speed doesn't matter at all as it won't be accessed that frequently...[08:55:38] <DanaG> All the spare 2.5 drives I have make annoying clicky noises, that are more annoying than the Samsung drives.[08:56:33] <DanaG> I'd rather figure out why the USB stick is so slow in Solaris. Slow enough to take at least 10 minutes to boot.[08:56:46] <DanaG> And when I did a "text" install, it was impossible to log into the installed system.[08:57:49] <miine> did you disable the broadcom nic in bios?[08:58:31] *** gea__ has joined #openindiana[08:59:07] <DanaG> Haven't tried USB boot since before I put in the Intel card (today).[08:59:15] <DanaG> Is that what's doing it? The BGE driver?[08:59:55] <miine> you have either to disable the broadcom nic in bios or disable the bge driver via grub. I think bios is easier...[09:00:09] <DanaG> I got the IPMI card, as well.[09:00:24] <miine> the not updates bge driver does ten-thousands of interrupts per second, no wonder the system gets slow.[09:00:31] <miine> but at least it boots...[09:00:45] <DanaG> Would a slow target drive have caused the Text install to fail to create my user?[09:00:57] <DanaG> Or should I use what I know works (the graphics install)?[09:01:21] <miine> oh. let me think about it ...[09:01:34] <miine> of course use what you know what works :-)[09:01:45] <DanaG> yeah, I realized that just after I asked.[09:01:51] <DanaG> I wonder why the text one was broken like that.[09:02:35] <miine> I guess you'll never find out...[09:03:51] <DanaG> I wish AMD would make an IPMI chip with an ES1000, so we could at least get non-crappy 2D.[09:04:16] <miine> you need IPMI for what?[09:04:19] *** Avantopia has left #openindiana[09:04:40] <DanaG> So I don't have to lug out my old CRT from my closet to control the BIOS, and such.[09:05:02] <miine> how often do you change that settings?[09:05:26] <DanaG> Same is true of the openindiana boot menu.[09:05:31] <DanaG> Does it default to SSH?[09:05:53] <miine> you can change the grub menu at your will...[09:06:39] <miine> if you need a boot live medium with ssh as default, build yourself one with the distribution constructor.[09:06:42] <DanaG> I wish HP would fix their ACPI tables to specify the correct memory address.[09:06:56] <DanaG> Anyway, I already have the card, so no going back now.[09:07:19] <DanaG> In hindsight, yes, it is somewhat overkill.[09:07:26] <miine> the changes needed to do that are that obvious, it's done in 10 minutes. just building the media takes 1-2 hours depending on the cpu speed...[09:08:39] <alanc> ssh is off by default on the live media since the security implications of shipping a bootable system with a remote login open to the world with a known root password are staggering[09:09:06] <DanaG> Hmm, here's one idea: make it so you can hook up a keyboard, but no monitor, and hold a key to get SSH.[09:09:22] <DanaG> So you'd hold some magic combination, and it'd enable SSH blindly.[09:10:02] <miine> why that fuss if you can build your own live cd that easily ?[09:10:50] <miine> it also makes sense to have the updated bge driver on it anyway.[09:12:38] <alanc> it's a rare x86 system that you can switch boot devices without a keyboard/monitor or some sort of remote console hooked up[09:13:19] *** johnny_N30 has quit IRC[09:13:39] *** nikolam has joined #openindiana[09:17:48] <DanaG> vPro on desktops can boot from arbitrary device types on command -- CD or HDD.[09:18:00] <nikolam> hi, does deleting snapshots is working on /dev-il 151b? Here is not.zpool is Version 14 and zfs is version 3 (not changed from 2009.06, 111b)[09:19:38] <nikolam> seems like backward compatibility for zfs does not count snapshot destroy on older zpools[09:28:13] <DanaG> hmm, with broadcom disabled, bnx is still loaded.[09:30:57] <miine> DanaG: bnx? you mean bge ?[09:31:29] <DanaG> Nope. dmesg shows exactly one line:[09:31:47] <miine> what does ifconfig say?[09:31:50] <DanaG> bnx: [....] NOTICE: Broadcom NetXtreme II Gigabit Ethernet Driver 6.0.1[09:32:05] <DanaG> ifconfig doesn't list interfaces like Linux does...[09:32:19] <DeanoC> ipadm show-addr[09:32:20] <miine> sure it does. ifconfig -a ?[09:32:34] <DanaG> ah, -a... one of those ones that doesn't say what it does.[09:32:46] <DeanoC> ifconfig is largely legacy (except for a few advanced things)[09:32:56] <DanaG> Okay, so the driver loaded, but didn't bind to anything.[09:33:26] <miine> if the interface doesn't show up, you're fine...[09:33:48] <miine> I also guess it booted faster :-)[09:33:50] <DeanoC> dladm show-phys will show you network dvices[09:34:20] <DanaG> The install is still pretty slow. 100 %w and %b on iostat -CX 1[09:34:27] <DanaG> '1' being an interval.[09:34:43] <miine> installing is always slow...[09:34:53] <DanaG> I'll have to re-judge once I'm booted from the installed system.[09:35:29] <DanaG> Heck, if OpenSolaris is stable enough, I can muck around with the Linux guest all I want.[09:35:37] <DanaG> And put the VM disk itself on the data pool.[09:36:38] <DanaG> Too bad it doesn't do IOMMU... then I could give the VM the GPU.[09:36:55] <DanaG> But that'd also need a BIOS option to boot from add-in, but leave onboard enabled.[09:37:54] <DanaG> I wonder what'd happen if I put the IPMI card in some random low-profile PCIe 1x computer.[09:38:49] <DanaG> Which would be faster: VBox -> iscsi -> Solaris pool, or VBox -> VBox hard disk on pool?[09:40:00] *** bens1 has joined #openindiana[09:40:05] <miine> DanaG: what would be easier?[09:40:15] <DanaG> The latter, of course. =P[09:40:28] <miine> DanaG: first get it running, then you can speed up if needed later...[09:40:40] <DanaG> I'm also not sure whether I'd go over Samba to the torrent dirs, or over NFS. Do Samba and NFS lock each other?[09:40:56] <DanaG> Everything else would be using Samba.[09:41:20] <miine> DanaG: you don't need samba. use the build-in kernel CIFS server. its about 50%-100% faster...[09:41:52] <miine> DanaG: use samba if you want to share something to the world and install samba in a zone.[09:42:41] <DanaG> ah. I intend to have a share for my stuff, and a share for parents' stuff.[09:42:49] <DanaG> We don't need to access each other's stuff.[09:43:36] <miine> DanaG: you can share via cifs by setting the zfs dataset's parameter to do so (as you would doing for nfs...)[09:43:47] <DanaG> Sweet.[09:43:49] <miine> DanaG: so you can have any number of cifs shares...[09:44:06] <DanaG> I did get that portion... one dataset for me, one dataset for them.[09:44:25] <miine> but be prepared that zfs ACLs can be tricky. so stay at the unix permissions to have it running...[09:44:26] <DanaG> (And once I go on my own, I'll build another one of these microservers, if I can convince them to like having a NAS).[09:44:54] <DanaG> What they currently have: an iMac, and a samba share that my dad uses to access the stuff under that one user.[09:47:19] <DanaG> ugh, I suck at making myself go to bed when I need to.[09:55:01] *** DanaG has quit IRC[09:56:44] *** Naresh has quit IRC[10:09:49] *** SH0x has quit IRC[10:13:58] *** SH0x has joined #openindiana[10:17:03] *** bens1 has quit IRC[10:42:28] *** nikolam has quit IRC[10:43:28] *** nikolam has joined #openindiana[10:55:22] <nikolam> Should not uname -a spit out also Illumos release version? I suppose Illumos did not releases one yet but there are 1.0, 1.1 designations on Illumos bug reporting already.[11:14:11] *** Naresh has joined #openindiana[11:17:17] *** gea__ has quit IRC[11:22:59] *** Naresh has quit IRC[11:24:16] *** Naresh has joined #openindiana[11:26:08] *** ThomasB2k has quit IRC[11:39:32] *** tsukasa has quit IRC[11:42:58] *** ivo_ has joined #openindiana[11:43:35] *** merzo has joined #openindiana[11:44:36] *** tsukasa has joined #openindiana[12:00:47] *** xmikus01 has joined #openindiana[12:52:59] *** robinbowes has quit IRC[12:55:37] *** forquare has joined #openindiana[13:01:12] *** forquare has quit IRC[13:03:01] *** Oriona has quit IRC[13:10:18] <nikolam> Next thing I have often in mind is: Where is sort of table or something that explains what are actually Program packages VERSIONS in IPS instead just of entire IPS release?[13:14:43] *** Oriona has joined #openindiana[13:19:47] <nikolam> Anyone working on some locales/languages for OpenIndiana ubder Illumos? Is there a skeleton for filling to make language support per-language?[13:21:38] *** Oriona has quit IRC[13:25:04] *** freedomrun has quit IRC[13:34:59] *** dekar has quit IRC[13:36:23] *** xmikus01 has quit IRC[13:36:48] *** DrLou has joined #openindiana[13:36:48] *** ChanServ sets mode: +o DrLou[13:38:46] *** Oriona has joined #openindiana[13:39:38] *** sergefonville has joined #openindiana[13:39:52] *** dekar has joined #openindiana[13:46:58] *** xmikus01 has joined #openindiana[13:52:40] *** Oriona has quit IRC[13:56:40] *** Oriona has joined #openindiana[14:02:24] *** robinbowes has joined #openindiana[14:07:50] *** raichoo has joined #openindiana[14:09:24] <raichoo> hi folks[14:09:55] *** nikolam has quit IRC[14:21:49] <sergefonville> hi raichoo[14:22:17] *** bens1 has joined #openindiana[14:26:34] <DrLou> morning, all...[14:27:38] <RoyK> mrnng[14:29:41] <sergefonville> how are we today?[14:47:32] <sergefonville> how do I increase the max open files?[14:51:26] <tsoome> still havent figured out?:P[14:51:44] <tsoome> look on output of prctl $$[14:52:31] <tsoome> those limits you see, can be set in /etc/project (for static setup) or changed online for live change[14:53:12] *** bens1 has quit IRC[14:53:57] <tsoome> you also wanna read http://download.oracle.com/docs/cd/E19963-01/html/821-1460/resource.html[14:54:02] *** InTheWings has joined #openindiana[14:55:19] <tsoome> also note the 32bit stdio interface got smaller limits than 64bit stdio, so building 64bit app will "unlock" more resources for your app.[14:55:48] *** ksx4system has joined #openindiana[14:59:35] <sergefonville> I found something about /etc/system rlimit_fd[14:59:43] <sergefonville> but not quite sure if that is it[15:00:13] <sergefonville> ulimit -n gives 256[15:02:44] <tsoome> there are still some leftovers from old knobs, all static and change needs reboot. the proper way today is using /etc/project (and resource controls in zonecfg), because all those can be managed on the fly and threfore are more flexible.[15:04:20] <tsoome> man rctladm(1M) and prctl(1)[15:14:47] <sergefonville> do I need to change or add a project?[15:15:39] <tsoome> you can do any way - either create new one or change default[15:16:40] *** McBofh has quit IRC[15:17:10] <sergefonville> how do I tell the system when which project is used?[15:17:16] <tsoome> projects command will tell you which ones you have. the effect is the same as global settings versus local ones, as changing the default project will change all apps and that may not be always a good idea:)[15:18:40] <sergefonville> so, how do I create one specifcally for a certain process?[15:22:08] *** McBofh has joined #openindiana[15:22:38] <sergefonville> for example I found a doc from IBM for websphere that says to increase the process.max-file-descriptor to 10,000 for group.mqm[15:23:16] <tsoome> you can bind the project to user or group, make project an default project for user if needed (usually done for oracle user for example), and if the project is not default, you can assing the app to project by newtask command.[15:25:07] <sergefonville> I am trying to increase the max open files for nginx[15:25:21] <sergefonville> I added the setting and it is in /etc/projects[15:27:49] <sergefonville> prctl -t basic -n process.max-file-descriptor $(pgrep nginx) givees 256[15:28:00] <tsoome> if the service is started by smf, you can set project name as described in man smf_method[15:28:20] <sergefonville> projects -l default[15:28:20] <sergefonville> default[15:28:20] <sergefonville> projid : 3[15:28:20] <sergefonville> comment: ""[15:28:20] <sergefonville> users : (none)[15:28:21] <sergefonville> groups : (none)[15:28:21] <sergefonville> attribs: process.max-file-descriptor=(basic,1024,deny)[15:29:33] <tsoome> the project file is just static setup used at login or application start, if the app is already running, you need to either restart it or use rctladm[15:32:12] <tsoome> i guess the nginx is started from smf? then create project for it and set project property for the service and restart it, that way you dont have to change default[15:33:32] <tsoome> also that would make it safer for later management, you can change that project on the fly without having to worry how the change will affect other apps.[15:44:39] <sergefonville> what should I name the project and where do I reference it?[15:45:25] <tsoome> name is whatever you like:P[15:46:33] <tsoome> man smf_method; you need to set project in method contex with svccfg[15:46:37] *** hsp has joined #openindiana[15:50:24] <tsoome> http://www.c0t0d0s0.org/archives/4209-Less-known-Solaris-Features-Resource-Management-Part-6-Resource-Management-and-SMF.html[15:50:28] <tsoome> some examples:)[15:53:08] *** freedomrun has joined #openindiana[15:53:41] <sergefonville> slowly I'm getting there :P[15:53:50] <sergefonville> thank you very much so far[15:54:12] <tsoome> it will take a bit, but its quite nice interface once you got the idea:)[15:55:13] <sergefonville> if I want to add the project to the manifest, can I add an attribute to the exec_method[15:55:51] <sergefonville> or do I need to add a property_group start for that and add an propval for project?[15:56:06] <tsoome> there is an example et the end of the blog entry[15:56:14] <tsoome> s/et/at/[16:02:43] <sergefonville> it still is 256 :([16:02:59] <tsoome> ?[16:03:13] <tsoome> is the project name correct one?[16:09:32] <sergefonville> it didn't apply the project[16:09:33] <sergefonville> ps -o pid,project,args -ef | grep nginx[16:09:36] <sergefonville> gives system[16:09:37] <tsoome> what user was used to start the nginx?[16:10:21] <tsoome> the user should be listed in projects users list or you cant use project as that user[16:10:29] <tsoome> newtask: user "tsoome" is not a member of project "test"[16:12:14] <sergefonville> but nginx is first started as root and then forked as webservd[16:12:22] <sergefonville> do I need to specify them both then?[16:13:43] <tsoome> maybe, i never checked about that[16:14:21] <tsoome> anyhow, I did create test project: test:100:Test:tsoome::process.max-file-descriptor=(basic,10000,deny)[16:14:29] <tsoome> newtask -p test bash[16:14:45] <tsoome> and prctl -n process.max-file-descriptor $$ does confirm i have 10k on basic[16:14:56] <tsoome> basic 10,0K - deny[16:15:41] <tsoome> once you can verify like that, the second step is to fix service startup to use new project and off you go:)[16:18:23] <tsoome> in that ibm example you were referring, they assigned the project directly with whole mqm group and hence dont need to list mqm group in that project group list, its like primary group in passwd versus group member setup in group file[16:22:13] <sergefonville> prctl -t basic -n process.max-file-descriptor $(pgrep nginx)[16:22:13] <sergefonville> process: 2002: /usr/sbin/nginx[16:22:13] <sergefonville> NAME PRIVILEGE VALUE FLAG ACTION RECIPIENT[16:22:13] <sergefonville> process.max-file-descriptor[16:22:13] <sergefonville> basic 1.02K - deny 2002[16:22:19] <sergefonville> so solved :D[16:22:34] <sergefonville> thank you tsoome, your awesome!![16:22:41] <tsoome> you are welcome.[16:23:00] <tsoome> what was wrong before?[16:24:12] <sergefonville> the actual problem seems to stil be there :([16:24:25] <sergefonville> if I run /usr/bin/amd64/ab -n 8192 -c 256 http://192.168.1.1/[16:24:34] <sergefonville> I get socket: Too many open files (24)[16:25:06] <tsoome> pfiles will list you all open files[16:25:15] <tsoome> so you can check how many it got...[16:25:38] *** McBofh has quit IRC[16:27:25] <sergefonville> 24...[16:27:27] *** mnaser has joined #openindiana[16:27:55] <sergefonville> does every socket connection count for a file descriptor?[16:28:00] <tsoome> yes[16:28:18] <sergefonville> so even though nginx has 1025 ab does not[16:28:33] <sergefonville> and therefore I cannot test with more than 224[16:28:54] <tsoome> ah, the error was thrown by ab and not nginx?[16:29:17] <tsoome> so run ab from your new profile (newtask -p profile ab ....[16:29:20] <sergefonville> well at first it was both. since I ran ab also from an XP system[16:30:53] <sergefonville> hmmm, th higher the concurrency the lower requests per second[16:31:56] <tsoome> how many workers you have?[16:32:38] *** mnaser has quit IRC[16:32:53] <sergefonville> 10[16:34:42] <tsoome> http://nginx.2469901.n2.nabble.com/Tuning-workers-and-connections-td3192878.html[16:34:44] <tsoome> ?[16:35:50] <tsoome> basically if you have too few workers, then each worker will get an connection queue and once that queue is "too long", your response latency will grow[16:36:42] <tsoome> so you need to create some sort of balance there[16:36:45] *** mnaser has joined #openindiana[16:45:00] <sergefonville> with low amounts of concurrent en total request I get close to 10,000 per second[16:46:45] <sergefonville> s/ en / and /[17:03:01] *** sergefonville has left #openindiana[17:03:07] *** sergefonville has joined #openindiana[17:09:40] <sergefonville> how do I know if the performance is impacted by disk IO?[17:10:05] <sergefonville> sicne the difference is only this varying on localhost[17:10:23] <sergefonville> over the network its usuall steady at 250 req/s[17:10:26] <sergefonville> since*[17:10:44] *** floogy has joined #openindiana[17:17:43] *** dekar_ has joined #openindiana[17:21:07] *** dekar has quit IRC[17:23:18] *** merzo has quit IRC[17:24:01] *** miine has quit IRC[17:25:49] *** sergefonville has quit IRC[17:40:42] *** miine has joined #openindiana[17:51:34] *** Botanic_ has quit IRC[17:57:11] *** sergefonville has joined #openindiana[18:03:10] *** raichoo has quit IRC[18:19:15] *** xmikus01 has quit IRC[18:20:09] *** nikolam has joined #openindiana[18:26:55] *** movement has quit IRC[18:33:55] *** nikolam has quit IRC[18:34:33] *** freedomrun_ has joined #openindiana[18:35:13] *** freedomrun has quit IRC[18:35:51] *** ShadowHntr has joined #openindiana[18:35:51] *** ShadowHntr has quit IRC[18:35:51] *** ShadowHntr has joined #openindiana[18:38:23] *** movement has joined #openindiana[18:41:08] *** nikolam has joined #openindiana[18:45:34] *** DrLou has quit IRC[18:45:35] <sergefonville> if I add priviliges to method_credential are they automatically added to /user_attr when I import the manifest?[18:46:02] *** dekar__ has joined #openindiana[18:49:31] *** dekar_ has quit IRC[18:51:46] *** dekar__ has quit IRC[18:56:02] *** dekar__ has joined #openindiana[18:57:56] *** ThomasB2k has joined #openindiana[19:02:43] *** floogy has quit IRC[19:08:34] <sergefonville> no, it does not, priviliges are separate from what's in /etc/user_attr[19:08:59] <sergefonville> I do have another question....[19:09:02] <sergefonville> regarding SFE[19:09:19] <sergefonville> how do I suggest updates to a spec file?[19:10:29] <Triskelios> send a patch to the mailing list[19:10:35] *** movement has quit IRC[19:22:05] *** movement has joined #openindiana[19:26:59] *** freedomrun has joined #openindiana[19:27:19] *** freedomrun_ has quit IRC[19:28:48] *** ivo_ has quit IRC[19:53:56] *** ShadowHntr has quit IRC[19:58:31] *** Naresh` has joined #openindiana[20:00:00] *** Naresh has quit IRC[20:44:43] *** dekar__ has quit IRC[20:49:26] *** dekar__ has joined #openindiana[20:59:01] *** redgone has joined #openindiana[21:18:08] *** tsukasa has quit IRC[21:21:04] *** tsukasa has joined #openindiana[21:22:59] *** xmikus01 has joined #openindiana[21:26:22] *** sivanov_ has joined #openindiana[21:27:20] <sivanov_> finally yeah, iozone shows some 110MB/s read spead on windows, on a cifs share[21:36:27] <dkeav> thats pretty good over cifs[21:37:30] <sivanov_> installed today iozone on windows, upped tcp window sizes on wondows and oi to 1024KB[21:37:54] <blues> can i flash this intel controller to IT mode from Solaris, or do i have to use a windows box?[21:38:20] <sivanov_> they had like... dos .exe[21:39:56] <sivanov_> found this supermicro H8DII+-F motherboard on ebay[21:40:00] * sivanov_ wants[21:40:19] <dkeav> blues: you have to use dos[21:40:26] <sivanov_> but $364.99 hard to justify for home use[21:41:01] <blues> dkeav: so i've got to install it into a windows box to flash it[21:41:14] <sivanov_> 16 DIMMs, 4 pci-e 2.0, 2 pci-x for older cheaper NIC nad FC HBA[21:41:39] <sivanov_> and 8 sas ports[21:41:48] <RoyK> sivanov_: they are good, though, those mobos[21:42:11] *** scarcry has quit IRC[21:42:20] <dkeav> blues: i usually just use a win98 usb stick i made up for flashing bios's[21:42:43] <RoyK> X8DTH is even better ;)[21:42:46] <sivanov_> will it see the memory in both banks with only one opteron?[21:42:58] <RoyK> no[21:43:03] <RoyK> it uses NUMA[21:43:06] <dkeav> i used hp's usb utility to put the bootdisks.com win98 image on a 128mb flash drive, then i just mount it under solaris or linux or whatever and throw the flash utility and rom on there and boot it[21:43:28] <blues> curiously, i wasn't paying attention and went ahead and let OI boot...and it sees all teh drives on the controller[21:44:34] <dkeav> when you booted did you go into the controllers rom and see what settings it currently had?[21:44:44] <blues> i just rebooted to see[21:45:17] <blues> lsi cor config utility, MPT firmware revision 1.26.00.00-IR[21:45:23] <sivanov_> well, i can put 32 mb in one bank, using 4gb dimms, but buying second opteron and using 2gb dimms would be probable chaper[21:45:32] <sivanov_> *gb[21:48:04] <sivanov_> i think, x8dth will make great fc target[21:48:55] <sivanov_> those supermicros are rock solid[21:49:24] <kdavy> sivanov_: i have three units based on x8dth, with qlogic 2460 and 2462 fc cards and LSI 9200 SAS - highly recommend[21:50:27] <kdavy> only thing that sucks about it is, if you get a 2U enclosure for it good luck finding low profile brackets for the QLogics - i had to make mine with a Dremel kit[21:51:28] <dkeav> blues: yea you have the IR firmware, you want the IT[21:52:11] <blues> yeah.. working on making a bootable usb stick i can install from.[21:52:21] <sivanov_> but on them one can have 40 SAS ports cheap and without using an expander[21:52:38] <dkeav> blues: do you have the HP utility?[21:52:57] <kdavy> sivanov_: yeah, 40 sas ports easy[21:53:20] <blues> somewhre. We are in the process of selling our house..so a ton of my stuff is in storage. I may have to find / redownload, blah blah[21:53:30] <kdavy> mine only have 8 ports each, since i use FC for disks as well[21:55:13] *** blues has quit IRC[21:55:52] <sivanov_> iirc i ve seen maximum 35 3.5 sas drives in one 4U supermicro chassis[21:56:21] <sivanov_> the one that has supplementary drive basket on back[21:56:34] <RoyK> sivanov_: 36[21:56:35] *** blues has joined #openindiana[21:56:41] <sivanov_> *35[21:56:43] <RoyK> or was that 35...[21:56:50] <kdavy> there's a chassis with 45 drives in 4u, but no space for the motherboard[21:56:52] <RoyK> 45 in a JBOD[21:56:54] <sivanov_> *36 lol[21:57:11] <kdavy> also they have one with 72 or 88 2.5" drives in 4U - i've been eyeing that one[21:57:19] <RoyK> dkeav: I know, I have a couple of them ;)[21:58:04] <sivanov_> can solaris fully saturate two 8G FC?[21:58:09] <dkeav> drives for flashing bios?[21:58:25] <dkeav> yea pretty handy[21:58:27] <tsoome> its more about if you machine can:P[21:59:04] <tsoome> your*[21:59:19] <blues> formatting stick now[21:59:27] <sivanov_> well, at work my machine can 180 write/350 read but over iscsi its miserable 55MB/s[21:59:28] <kdavy> i've saturated two 4G FC with a single Xeon 5620, and cpu usage wasn't high at all[21:59:34] <sivanov_> with 4 links[21:59:43] <sivanov_> iḿ yet to figure it why[21:59:53] <tsoome> what is behind iscsi?[22:00:07] <kdavy> sivanov_: jumbo frames? tcp window size adjusted?[22:00:16] <sivanov_> yep[22:00:17] * kdavy isn't an iscsi expert[22:00:17] <tsoome> the storage, i mean[22:00:34] <tsoome> zvol?[22:00:40] <sivanov_> yes[22:00:43] <tsoome> well.[22:00:46] <sivanov_> wcd=false[22:00:55] <sivanov_> 128kb record size[22:00:59] <tsoome> what are the clients?[22:01:10] <sivanov_> esx 4.1u1[22:01:44] <kdavy> what's your multipath option set to on storage?[22:01:47] <tsoome> well, you can set sync=disabled on zvol for an test and see what will happen.[22:01:50] <kdavy> on esxi that is[22:01:50] <sivanov_> rouund robun on, number of IOs over one channel tuned[22:02:22] <sivanov_> well, i get 55MB/s read[22:02:33] <tsoome> ah, that 55MB/s is for reads or writes?[22:02:38] <sivanov_> both with fixed path and round robin[22:02:43] <sivanov_> both[22:03:11] <sivanov_> iirc write is even higher[22:03:47] <tsoome> well, reads are not affected by wcd or sync, that game is all for writes[22:04:41] <sivanov_> dont have any spare server with 4 porrts for iscsi, to try another client like windows server 2k8r2 or linux[22:05:19] <sivanov_> out of nics[22:05:23] <tsoome> have you tested with ftp or even scp, what normal file trasfer perfomance you see?[22:05:40] <sivanov_> tested with cifs[22:05:43] *** dekar__ has quit IRC[22:05:45] <tsoome> normal = without any iscsi or other filesystem overhead[22:05:51] <sivanov_> basicly, same read speed[22:06:56] <sivanov_> besides, nice intel nics on iscsi side of server[22:08:23] <tsoome> without any tuning you should be able to fill the link....[22:08:42] <sivanov_> on the other hand... in ESX i have two descrete NICs for VMs and console, and 4 built in ports, broadcom based[22:09:04] <sivanov_> esx is hp 380 g6[22:10:15] <tsoome> anyhow - test first with very simple apps, anything which is not tied too tightly with lower layers of storage - reading cached data or similar[22:10:29] *** dekar__ has joined #openindiana[22:10:55][22:11:34] <tsoome> once you get decent result there, its point to check with more complicated things as cifs/nfs/iscsi...[22:12:40] <sivanov_> i run one simple test; iozone -s10g -r1m -i 0 -i 1 -i2 -t 8 -k64 -w -F 1 2 3 4 5 6 7 8[22:13:48] <sivanov_> it saturates IO from local ESX datastore when i run it in centos vm[22:14:01] <sivanov_> but not iscsi[22:14:02] <tsoome> well, if there is some kind of issue at network level, the iozone wont really tell you anything:)[22:14:58] <tsoome> most probaly the vm did just fill up the command queue on iscsi target in esx and thats it:D[22:15:14] <tsoome> probably*[22:15:16] <sivanov_> well, i dont see any place for network issues: everything is connected to one cisco 3750, separated from remaining network[22:15:17] *** gea__ has joined #openindiana[22:15:49] <tsoome> thats why im suggesting to test with something simple - ftp, iperf etc...[22:15:52] <sivanov_> cat 6 casables aroound 4 m in length in one serverroom[22:16:09] <sivanov_> iperf gives wire speed[22:16:36] <tsoome> ftp read should give you same[22:16:56] <tsoome> at least on second try when the file is all in arc[22:17:12] <sivanov_> but i cant make use of 4 links with ftp[22:17:33] <tsoome> you can test one by one[22:18:22] <tsoome> you arent using some aggregation?[22:18:30] <sivanov_> no[22:18:41] <sivanov_> aggregation does not give anyting[22:18:48] <tsoome> yep.[22:18:52] <tsoome> just checking[22:18:54] <tsoome> :)[22:19:06] <sivanov_> and ESX does not support mc/s[22:19:12] <tsoome> you did test iperf on both direction btw?[22:19:18] <sivanov_> yes[22:19:39] <tsoome> hm, thats suggesting the nw layer is healthy...[22:20:05] <tsoome> if you mount some dataset with nfs and try reading from nfs, also slow?[22:20:41] <tsoome> if cifs and iscsi were, the nfs should be affected same way....[22:21:43] <sivanov_> cant experiment much on vmware: the way iscsi is configured involve binding the phys nics with vmkernel nics on-to-one and adding them then to iscsi stack[22:22:09] <tsoome> but with another host?[22:22:16] <sivanov_> cant configure more noeworks on those physical links[22:23:06] <tsoome> well, you need to start from somewhere:)[22:23:28] <sivanov_> other host plugged in this switch is centos host. however i saved iscsi testing with it on monday[22:23:36] <sivanov_> hoping find more nics[22:23:45] <tsoome> if you can repeat the same effect on another host, it would be some start. or if you can confirm, with another host you get decent speed[22:24:00] <sivanov_> wire speed with ipefs, everyting else untested[22:24:49] <tsoome> local io on filesystem is all ok i suppose?[22:24:50] <sivanov_> from access network to windows 7 host via cifs is 40-50mb/s[22:25:21] <sivanov_> 180MB/s read 350MB/s write[22:25:30] <sivanov_> local[22:25:40] <sivanov_> *read=write[22:25:53] <sivanov_> 180 write 350 read :)[22:26:48] <sivanov_> could have been better if not the crappy serveraid 8k and the backplane that does not work with sata2[22:27:34] <sivanov_> also raid10, so IOPS should not be a bottleneck[22:28:27] <sivanov_> so... even dont know where to look[22:28:47] *** axisys has joined #openindiana[22:29:31] <tsoome> getting file \tsoome\solaris\osol-0906-x86.iso of size 709871616 as osol-0906-x86.iso (11256.0 KiloBytes/sec) (average 11256.0 KiloBytes/sec)[22:29:38] <tsoome> thats from fast ethernet only[22:30:09] <tsoome> but cifs server is s11 express, pool is just mirrored 146GB 10KRPM scsi disks[22:30:26] <sivanov_> will probably take one NIC in oi server out and put it info linix host to see how linux will work over two links[22:31:00] <sivanov_> hmm maybe, even three links[22:31:22] <sivanov_> just neeed correct lines for /etc/multipath.conf[22:35:22] <sivanov_> btw, does changing LUN block size in stmfadm from 512 to 4096 gives anyting to the speed?[22:35:55] <tsoome> never tested.[22:36:26] <tsoome> are you using jumbo frames btw?[22:36:34] <sivanov_> yes[22:36:39] <tsoome> 9k?[22:36:44] <sivanov_> yep[22:37:05] <tsoome> well, you can start from there then, test with 1500....[22:37:29] <tsoome> maybe there is some issue with jumbos[22:38:22] <sivanov_> the 1500 is not what i want: the original plan was to connect an infortrend raid shelf to ESX @decent speed, and 1500 gives very poor performance in esx[22:39:20][22:39:37] <tsoome> well, thats not the point, what you need to verify is if thats related to the issue[22:40:38] <tsoome> if it isnt, you can just turn it back. if you get more perfomance with 1500, you can set ip larger with smaller steps[22:40:46] <tsoome> s/ip/it/[22:41:16] <sivanov_> on the other hand, cifs performance on other interface sucks and there no jumbos[22:41:41] *** Naresh`` has joined #openindiana[22:41:53] <tsoome> well, thats good point:)[22:42:12] <tsoome> other interface is same type and is using same switch?[22:43:15] *** Naresh` has quit IRC[22:44:34] <sivanov_> iscsi is e1000g, normal network is bnx[22:46:54] <sivanov_> same switch model, 3750 just two switches between my machine and oi server: switch in server room, switch in sysadmins room[22:46:54] <RoyK> tsoome: isn't 9k the usual size for jumboframes?[22:47:13] <RoyK> for 1Gbps at least[22:47:19] <tsoome> so is 1500 for ethernet, but you can change it[22:47:42] <tsoome> for example, solaris fcoe tells you need minimum 2500[22:48:21] *** Naresh``` has joined #openindiana[22:49:12] <sivanov_> i like expandind windows system disk which sits on solaris over FC :)[22:49:21] <sivanov_> *expanding[22:49:40] <tsoome> damn, cant remember my another server lom ip and stupid host is powered off:([22:49:49] *** Naresh`` has quit IRC[22:51:21] <sivanov_> hmm 2 days till dances with dragons left :)[22:51:45] <tsoome> ah, found it, now i can test cifs and friends over gigabit:D[22:57:30][22:58:40] <sivanov_> zpool iostat shows some 410MB/s writes[22:58:48] <sivanov_> but cpu usage is 55%[23:02:09] <tsoome> intrstat 1; mpstat 1[23:04:03] <sivanov_> about 700 ints to qlogic, 2000-6000 ints to aoc-sat2-mv8 controllers[23:05:09] <sivanov_> on which numbers to look in mpstat?[23:05:46] <tsoome> which cpu is most loaded and if it will be same cpu which is serving those interrupts...[23:06:27] <dkeav> anyone bootstrapped pkgsrc to oi_151 yet?[23:07:41] <sivanov_> two three controllers use cpu #2, none use #0[23:08:33] <sivanov_> is there way to rewire interrupts for qlt or marvell88sx??[23:09:26] <richlowe> "wait and let intrd do it, if it decides it'd be a good idea" would be one option.[23:10:45] <Meths> Does pkgrecv (or the destination repo) automatically take care of changing publishers during the process?[23:10:49] <sivanov_> as i look it apparently not going to[23:11:02] <sivanov_> and cpu #0 has 0 interrupts[23:12:59] <tsoome> it may have good reason;)[23:13:01] <blues> hmm[23:13:34] <blues> i'm attempting to use sasflash utility to flash the intel card... but i keep getting "FAILED TO INTIALIZE PAL"[23:13:37] <tsoome> intrd is perl script, you can look into it[23:15:32] <sivanov_> on #0 and #3 intr counters 2 times lowerr than on other cpus and ithr 100 times lower[23:16:17] <tsoome> hm, does solaris kernel smb do ipv6?[23:16:26] *** nikolam has quit IRC[23:18:49] <sivanov_> i wonder if i really should have made win7 zvol with recordsize 64k[23:19:08][23:19:15] *** nikolam has joined #openindiana[23:20:35] <tsoome> ld.so.1: smbclient: fatal: relocation error: file /usr/lib/libreadline.so.5: symbol tgetent: referenced symbol not found[23:20:38] <tsoome> cool.[23:24:57] <sivanov_> hmm physical memory usage 99% on windows while running iozone[23:25:40] <blues> well this is not cool.. i'm following the directions from this: http://www.servethehome.com/flashing-intel-sasuc8i-lsi-firmware-guide/ exactly, but no go[23:25:49] <sivanov_> looks like it desperately tries to cache random access over 40gb files[23:27:26] <tsoome> quite probably;)[23:30:34] *** McBofh has joined #openindiana[23:32:33] *** dekar__ has quit IRC[23:41:27] *** dekar__ has joined #openindiana[23:42:00] <sivanov_> hmm 160MB/s read in windows over FC4 and 134MB/sec re-read, and 50MB/s random read @1MB block[23:42:15] <sivanov_> why so low re-read speed??[23:42:18] *** ivo_ has joined #openindiana[23:43:40] <tsoome> run arcstat during test? so you can see what misses will do?[23:44:30] <sivanov_> test files are huge and dont fit in RAM[23:44:49] <sivanov_> 4*10GB files, 16GB ram[23:46:44] <sivanov_> 46MB/s random write[23:47:32] *** DanaG has joined #openindiana[23:47:45] <sivanov_> should run series of tests to understand what recordsize is best for windows[23:47:56] <DanaG> Say, is it a good idea to make a pool be named "home", so it mounts on /home? Or is that just asking for breakage?[23:48:10] <tsoome> DanaG: man automount[23:48:24] <tsoome> if you need to ask after that again....[23:49:08] *** sivanov_ has quit IRC[23:49:10] <tsoome> its not problem if you have configured your system properly;)[23:52:22] *** xmikus01 has quit IRC[23:53:25] <DanaG> The only oddity I can imagine is that things will fail badly if the home dir is not mounted.[23:53:35] <DanaG> How likely is that to happen (pool not mounted)?[23:53:55] <tsoome> oh well.[23:53:56] <tsoome> :P[23:54:03] <tsoome> read /etc/auto_master[23:56:39] <DanaG> hmm, maybe I'll just have to try it. Can you rename pools later?[23:56:44] <DanaG> er, lemme' google that.[23:57:05] <tsoome> sigh. well, just comment out /home line from /etc/auto_master .....[23:57:42] <tsoome> .oO howcome todays people cant read:P[23:57:49] <DanaG> Sorry, I just wasn't in OI at the moment -- the server, which is booted from a flash drive, is shutting down..... very very slowly.[23:58:25] <DanaG> Somehow, that flash drive plus OI is way slower than any other combination of OS and drive.[23:59:32] <DanaG> Turns out it's not the bge driver making it slow, after all... even with Intel card added and bge card disabled, it's still slow.