July 16, 2012 
< | 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 16, 2012 [00:14:48] *** haggle has joined #smartos[00:15:25] <haggle> hey, random question, has anyone had any success / failure at getting VRRP running in a smartos "joyent" brand zone ?[00:16:03] <haggle> I've got the vrrpd service running, and am about to hack the joyent brand file to add the vnic configuration options, just wondered (before I do) if there is a massive blocker??[00:22:23] *** noahmehl has joined #smartos[00:22:37] *** enmand has quit IRC[00:31:02] *** enmand has joined #smartos[00:33:04] *** Orpheus has joined #smartos[00:33:20] <Orpheus> yay i got the smartos/joyent stickers :D[00:33:38] *** Orpheus is now known as Orpher[00:33:50] *** Orpher is now known as Orpheus[00:34:45] *** mamash has left #smartos[00:46:36] *** noahmehl has quit IRC[00:46:58] *** knweiss has quit IRC[00:49:28] <benbangert> so no one has any idea how to set a hostname that sticks on reboots?[00:49:48] <benbangert> I mean, seems pretty lame that the docs fail to mention it, and sendmail and shit throws errors until /etc/hosts is fixed, but of course, that doesn't persist across restarts[00:56:02] *** bixu has joined #smartos[00:56:41] <haggle> benbangert: yeah it kinda sucks. I have a custom build of smartos with a hostname=<> parameter in the config file[00:56:56] <haggle> benangert: you could always drop an SMF manifest to set the hostname on boot ?[00:57:33] <benbangert> ah, how would I do that?[00:57:53] <haggle> create /opt/custom/smf[00:58:04] <haggle> any SMF manifest XML files in there will get pulled in on boot[00:58:08] <haggle> (and /opt lives on the zpool)[00:58:26] <haggle> so create a simple transient manifest that launches hostname foobar.domain[00:58:39] <haggle> http://wiki.smartos.org/display/DOC/Setting+a+static+hostname+at+boot+time[00:58:41] <haggle> there, in fact[00:59:01] <benbangert> ah![00:59:06] <benbangert> thought I went through every menu[00:59:11] <haggle> hehe[01:00:03] *** marsell has joined #smartos[01:01:22] <haggle> FYI: here is my patched identity-node method[01:01:22] <haggle> https://github.com/joshado/smartos-live/blob/master/overlay/generic/lib/svc/method/identity-node#L54[01:01:37] <haggle> that lets me configure hostname=foobar in my smartos config file[01:01:46] <benbangert> ahh, cool[01:01:53] <haggle> (and will fallback onto MAC address without it set, or when unconfigured)[01:01:54] <benbangert> that seems slightly easier than dropping a large XML file in :)[01:02:13] <haggle> yeah, but it's a sod to roll a smartos build unless you're set-up for it ;)[01:02:29] <benbangert> right, I can live without that[01:03:06] <haggle> Indeed. I've got quite a few patches now, so we have a build process that is actually really neat[01:03:14] <haggle> (all credit to Illumos / Joyent people!)[01:03:37] <rmustacc> haggle: Any reason you haven't submitted the patches?[01:03:47] <haggle> they're a bit single-use right now[01:03:54] <haggle> I'd love to - but, for example[01:04:01] <haggle> one patches sd.conf so that it disables power-save instructions[01:04:05] <haggle> (Dell firmwares go NUTS!!)[01:04:30] <haggle> and I've replaced the smartos config prompt thing with a node.js script that talks to a zookeeper server for configuration[01:04:49] <rmustacc> Interesting.[01:04:54] <haggle> the code is all on github, but it's pretty gnarly unless you know what it's doing[01:04:57] <rmustacc> What Dell Hardware are you running on?[01:05:01] <haggle> R610s[01:05:03] <haggle> err[01:05:05] <haggle> R620 (sorry)[01:05:11] <haggle> Dell H310s <-- not supported[01:05:17] <haggle> so flashed to the HBA firmware[01:05:23] <benbangert> this is working pretty nicely now, I should really make a usb key instead of booting via the ipmi mounted drive :)[01:05:24] <haggle> I tried bundling imr_sas solaris driver[01:05:25] <rmustacc> Ah you flashed the HBA firmware.[01:05:35] <haggle> this driver did _not_ go well :([01:05:38] <rmustacc> Okay, that answers that question.[01:05:41] <haggle> hehe[01:05:43] <haggle> kernel panic fun[01:05:49] <rmustacc> Oof.[01:05:57] <haggle> yeah, but only, AFAICT, when it was under load[01:06:04] <haggle> so I had a whole "whee, this is working" false sense of security[01:06:05] <haggle> then boom[01:06:31] <benbangert> is that the xeon e3 bug?[01:06:34] <haggle> I'm actually building a simple zone / hypervisor control page using Zookeeper as a config store[01:06:54] <haggle> benbangert: don't think so (not heard of that bug) but it's gone away now I've pulled that driver[01:07:17] <haggle> https://twitter.com/thomashaggett/status/223891503494348800/photo/1[01:07:19] <benbangert> it was this one, https://www.illumos.org/issues/1333[01:07:22] <haggle> noVNC ftw[01:07:52] <rmustacc> That one's been nicked for a while.[01:07:59] <haggle> no, it was a kernel panic in my case. Something about heap allocations[01:08:06] <benbangert> yea, but OI still doesn't include it. heh[01:08:07] <rmustacc> There are some other bugs with the x2apic, but you won't be hitting them.[01:08:08] <haggle> before debugging I pulled the driver and it all started working[01:08:12] <benbangert> its in their pre-release pkg[01:08:15] <rmustacc> Oh, it doesn't?[01:08:43] <benbangert> there were some docs I saw on how to switch to the pre-release pkg's so the fix could be pulled[01:09:03] <haggle> oh actually, whilst I'm on here, is there still an issue with perl in the build process ?[01:09:13] <rmustacc> Not that I know of.[01:09:15] <haggle> hmm[01:09:22] <rmustacc> But that doesn't mean there isn't one.[01:09:23] * haggle ssh in...[01:09:39] <haggle> [root@phy1-sw1-ash (ash) ~]# kstat[01:09:40] <haggle> Can't load '/usr/perl5/5.12/lib/i86pc-solaris-64int/auto/Sun/Solaris/Utils/Utils.so' for module Sun::Solaris::Utils: ld.so.1: perl: fatal: libintl.so.8: open failed: No such file or directory at /usr/perl5/5.12/lib/i86pc-solaris-64int/DynaLoader.pm line 200.[01:09:41] <haggle> at /usr/bin/kstat line 36[01:09:42] <haggle> Compilation failed in require at /usr/bin/kstat line 36.[01:09:43] <haggle> BEGIN failed--compilation aborted at /usr/bin/kstat line 36.[01:10:00] <rmustacc> Hmm. That's pretty weird.[01:10:02] <haggle> if I drop libintl.so.8 in, it all starts working again :([01:10:04] <rmustacc> What release is that from?[01:10:08] <haggle> the latest one[01:10:12] <benbangert> I've configured an additional NIC in usbkey/config, but its not marked INUSE, nor can I see it in ifconfig. is something else needed to bring additional NIC's online?[01:10:21] <haggle> I've got my patches on a git branch, I merge joyent's releases in[01:10:36] <haggle> I've had to do CC=/opt …[01:10:39] <haggle> in the perl makefile[01:10:48] <haggle> otherwise the build crashes[01:10:51] * haggle looks up the error...[01:10:57] <rmustacc> Did you update all of the illumos-extra branch as well?[01:11:12] <haggle> cw: error: mapping failed at or near arg '-fno-strict-aliasing'[01:11:16] <rmustacc> benbangert: Are you trying to have that nic be accessible from the global zone?[01:11:27] <benbangert> yea[01:11:34] <rmustacc> haggle: It almost sounds like you're behind a flag day or so.[01:11:35] <haggle> rmustacc: hmm, I didn't touch illumos-extra - I presumed that was pulled in during the ./configure step[01:11:37] <haggle> ah[01:11:45] <rmustacc> But I don't know so.[01:12:04] <rmustacc> ./configure won't update illumos-extra if it already exists.[01:12:12] <haggle> ah I do a fresh checkout for each build[01:12:31] <wesolows> gmake update will, however[01:12:33] <benbangert> wanted to do LACP in the gz[01:12:42] <haggle> benbangert: I added this, also[01:13:02] *** Orpheus has quit IRC[01:13:05] <benbangert> haggle: you got all my wants in your repo?? ;)[01:13:18] <haggle> heh sounds like it ;)[01:13:25] <haggle> https://github.com/joshado/smartos-live[01:13:27] <haggle> feel free :-p[01:13:51] <haggle> alas. I may have also cocked with /etc/issue, too:[01:13:52] <haggle> https://twitter.com/thomashaggett/status/223955593126354945[01:14:01] <wesolows> libintl.so.8 looks like it's GNU libintl, and comes from pkgsrc. It's not in the platform that I can see.[01:14:12] <haggle> wesolows: yeah, this makes sense - it works in zones[01:14:24] <haggle> and if I manually copy the binary into the GZ then it all starts working[01:14:31] <haggle> I was going to bodge it for the time being[01:14:32] <wesolows> that fits. But the platform perl shouldn't be trying to use that.[01:14:55] <benbangert> haggle: strange, the Downloads page seems to be missing the .iso ;)[01:15:04] <haggle> har har har. I don't use ISOs[01:15:17] <haggle> I just PXE boot the platform dirs[01:15:21] <benbangert> ah[01:15:47] <haggle> the build instructions are pretty sane, tbh[01:16:32] <haggle> wesolows: what would you suggest I do? I guess bodging binaries will work (technically) but it feels horribly dirty :([01:16:57] <benbangert> I'm partly surprised that ppl using SmartOS don't bond their NIC's[01:17:06] <wesolows> the platform files are correctly linked:[01:17:07] <wesolows> [root@bh1-build2 (bh1) /usr/perl5/5.12/lib/i86pc-solaris-64int/auto/Sun/Solaris/Utils]# ldd ./Utils.so libgen.so.1 => /usr/lib/libgen.so.1 libintl.so.1 => /usr/lib/libintl.so.1 libc.so.1 => /lib/libc.so.1 libm.so.2 => /lib/libm.so.2[01:17:25] <wesolows> note that it uses the platform libintl.so.1, not libintl.so.8.[01:17:35] <haggle> yeah, same for me[01:17:52] <haggle> libinto.so.8 just doesn't exist - yet perl Util (above) seems be linked against it[01:18:40] <wesolows> does elfdump -d show a DT_NEEDED entry for libintl.so.8?[01:19:06] <wesolows> I just noticed that there is a bug in this, but it shouldn't affect that: [4] RUNPATH 0x406 /usr/lib:/opt/local/lib[01:19:17] <wesolows> bad! bad bad bad! no /opt/local/lib in the platform! BAD![01:19:34] <rmustacc> haggle: You said that you were editing the Makefile?[01:19:37] <wesolows> (perl is one of 4 -extra components I didn't redo the build system for)[01:20:04] <haggle> rmustacc: yeah, feels like this is why it's broken (i.e. my own doing)[01:20:14] <haggle> but perl would fail to build (on a fresh box, etc)[01:20:14] <rmustacc> haggle: One question, what release were you doing the build on?[01:20:17] <rmustacc> e.g. what was the build machine?[01:20:23] <haggle> hmm[01:20:28] <haggle> latest IIRC[01:20:37] <rmustacc> There was a breif period of time where the smartos platform was broken for building perl.[01:20:39] <haggle> but I'm tempted to do an entirely fresh build, to double check[01:20:41] <haggle> ash right.[01:20:42] <haggle> *aah[01:20:51] <rmustacc> Are you sure you've merged everything in correctly?[01:21:22] <haggle> AFAICT. I'll double check and do a fresh re-build, though[01:21:35] <haggle> (not right now, mind, it's 20-past-midnight)[01:21:54] <haggle> but thanks for your time :)[01:21:55] <rmustacc> I'm going to quickly downlaod the latest image and see if our build of it has it linked correctly.[01:22:17] <haggle> thanks. I'm sure it's something I'm doing ;)[01:22:34] <haggle> Are you guys all at Joyent, then?[01:23:19] <rmustacc> wesolows and I are.[01:23:58] <haggle> awesome. I must say - great work. I've always liked opensolaris, etc, etc. SmartOS just had to be done :)[01:24:24] <haggle> just out of curiosity - if you can say - how many releases of the platform do you guys have in production at any one time?[01:24:39] <haggle> do you reboot nodes frequently, or have lots of releases running?[01:25:56] <rmustacc> I'm not part of the main operations for the cloud, but given the nature of the business, reboots don't happen often.[01:26:04] <haggle> thats what I figured, yea[01:26:38] <rmustacc> But that's why we try and keep the core both small and pretty well beaten on.[01:27:00] <haggle> makes a lot of sense :)[01:27:12] *** enmand has quit IRC[01:27:32] <haggle> and you presumably don't use LACP or anything like that? (or it's in an internal build?[01:27:33] <haggle> )[01:28:37] <bixu> Joyent usually schedules node reboots with customers, unless it's a dire emergency (this hasn't happened to me yet, and I run a lot of zones in JPC).[01:28:54] <benbangert> I would've thought it'd be easier to ensure all the bandwidth is used with LACP vs. hoping the VM's on each VNIC are balanced in usage[01:29:01] <bixu> Friendlier than the AWS "you are about to be rebooted" thing.[01:29:04] <haggle> hehe yes[01:29:31] <haggle> benbangert: it's less for the bandwidth, more for the link resilience for me…[01:29:50] <rmustacc> Honestly, I spend most of my time down in the kernel, so I'm not the best one to answer those kinds of things.[01:29:56] <haggle> hehe ok, sorry :)[01:30:16] <bixu> #joyent is maybe a better place to ask.[01:30:17] <haggle> just keen to know if there is a good reason why I shouldn't be using it ;)[01:30:20] <haggle> ah ok, thanks[01:30:23] <bixu> (I think that's a channel).[01:30:34] <bixu> haggle: Why you shouldn't be using what?[01:30:47] <haggle> vnics over an LACP trunk[01:30:53] <rmustacc> Well, to do LACP right you do have to have double the switches everywhere.[01:31:04] <haggle> I have a stack which does cross-stack LACP[01:31:12] <bixu> haggle: Do you *need* to trunk?[01:31:16] <haggle> not really, I Guess[01:31:38] <haggle> but AFAICT, I can kill a switch and not lose traffic[01:31:38] <bixu> I mean, most web apps are pushing mega, not giga, bits :)[01:31:43] <bixu> Ah.[01:31:47] <haggle> it's not for bandwidth, though[01:32:07] *** jelmd has quit IRC[01:32:22] <relling> it works as well as link aggregation works[01:32:38] <rmustacc> haggle: For what it's worth. I just fired up our latest iso and the kstat command is working / linked correctly.[01:32:47] <haggle> ok, thanks. I'll try a completely clean build[01:32:50] <relling> you might need to adjust the load-balance policy on both sides, though[01:33:21] <haggle> yeah, I've been tweaking those already[01:33:40] <relling> on the whole, link aggregation just sux[01:34:02] <haggle> ahha, this is useful. any particular reason why?[01:34:03] <bixu> rmustacc: Has there been any consideration of enabling NFS shares from within zones?[01:34:26] <benbangert> for me its bandwidth on LACP, cause I'm using it at home and I can push gigabits :)[01:34:30] <haggle> relling: i had it set up on opensolaris a while back and the e1000g driver used to duplicate traffic. which was fun[01:34:48] <haggle> heh. I had a hacked agent that transferred smartos dataset using bittorrent[01:34:49] <rmustacc> bixu: You mean having the zone be the NFS server?[01:34:50] <benbangert> I can see it not being a priority in SmartOS though, since its assumed these are web hosts[01:34:52] <haggle> that used to light up the switch[01:35:24] <relling> LACP is just the control protocol, and one of the most frequently misconfigured parts of link aggregation[01:35:40] <haggle> as in the balancing policy ?[01:35:43] <benbangert> right[01:37:26] <bixu> rmustacc: Yes.[01:38:33] *** jelmd has joined #smartos[01:38:44] <rmustacc> bixu: Well, it's something that would be nice to have, but our efforts are focused on other areas right now.[01:39:29] <bixu> rmustacc: No worries - just curious. Clearly NFS is a corner case for cloud computing.[01:40:20] <rmustacc> Some other folks in the illumos community started working on the problem, but I don't know where they are currently.[01:41:00] <haggle> right. calling it a night. thanks guys :)[01:41:25] *** haggle has quit IRC[01:43:04] <bixu> rmustacc: Maybe more attention will be paid to that now that you guys have made zone management so easy in SmartOS.[01:47:43] *** darjeeling has quit IRC[02:28:13] *** marsell has quit IRC[02:34:46] *** marsell has joined #smartos[02:35:09] *** noahmehl has joined #smartos[02:35:34] *** darjeeling has joined #smartos[02:40:00] *** guk has quit IRC[02:41:56] *** guk has joined #smartos[03:08:17] *** noahmehl_ has joined #smartos[03:08:47] *** noahmehl has quit IRC[03:08:48] *** noahmehl_ is now known as noahmehl[03:09:41] *** jamesd_ has quit IRC[03:31:15] *** kfr has quit IRC[04:25:03] *** denizr has joined #smartos[04:26:33] *** denizr has joined #smartos[04:43:58] *** denizr has quit IRC[04:45:32] *** denizr has joined #smartos[05:18:49] *** denizr has quit IRC[05:42:00] *** deirdres has joined #smartos[05:52:22] *** deirdres has quit IRC[06:07:46] *** deirdres has joined #smartos[06:57:01] *** deirdres has quit IRC[06:59:03] *** deirdres has joined #smartos[07:09:33] *** darjeeling has quit IRC[07:10:03] *** darjeeling has joined #smartos[07:39:39] *** darjeeling has quit IRC[07:42:42] *** darjeeling has joined #smartos[08:13:05] *** deirdres has quit IRC[08:13:59] *** sebasp has joined #smartos[08:20:29] *** Licenser has quit IRC[08:54:52] *** noahmehl has quit IRC[09:25:00] *** bens1 has joined #smartos[09:48:29] *** Licenser has joined #smartos[10:22:00] *** Licenser has quit IRC[10:25:15] *** Licenser has joined #smartos[10:54:51] <Licenser> trentster ping[11:09:36] *** darjeeling has quit IRC[11:22:25] <trentster> howdy mate.[11:23:04] *** trentster has quit IRC[11:25:33] *** bahamas10 has quit IRC[11:26:23] *** papertigers has quit IRC[11:26:40] *** bahamas10 has joined #smartos[11:26:49] *** papertigers has joined #smartos[11:30:45] *** trentster has joined #smartos[11:36:24] *** Licenser has quit IRC[12:14:17] *** darjeeling has joined #smartos[12:17:11] *** robinbowes has joined #smartos[12:17:15] <robinbowes> Morning all[12:18:08] <robinbowes> My zpool scrub is showing an error[12:18:33] <robinbowes> errors: Permanent errors have been detected in the following files:[12:18:34] <robinbowes> zones/dump:<0x1>[12:19:09] <robinbowes> I'm not too worried about this, but I'd like to clear the error[12:35:08] *** defa has joined #smartos[12:37:53] *** haggle has joined #smartos[12:38:25] <defa> Hello, I need some help with adding disk drives to a PXE booted SmartOS server… after adding drives the server boots into the smartos installer. After removing the disks everything is fine again...[12:40:55] <haggle> at a guess it's not finding the zpool[12:41:05] <haggle> can you get a shell on the box (i.e. bypass the installer)[12:41:07] <haggle> ?[12:44:06] *** max` has joined #smartos[12:45:34] <konobi> defa: you typed out each of the disks in question?[12:46:49] *** bens1 has quit IRC[12:46:53] <konobi> defa: also, what boot params are you supplying[12:54:06] <defa> haggle: I think so… need to find the root password on the smartos webpage...[12:55:50] <defa> konobi: regarding boot params, I'm sure what you mean… the PXE boot parameters?[12:57:34] <defa> konobi: used the tutorial on http://blog.smartcore.net.au/smartos-network-boot-with-pxe/ and adopted it to Mac OS X… so I guess I can pass the device with the /zones zpool somehow, right?[12:58:39] <haggle> on boot it should import all available zpools, and the system pool is marked with a hidden file[12:58:47] <haggle> so it sounds like there is a problem importing the zpool[12:58:59] <haggle> what I'd suggest is try manually importing the zpool and see what goes wrong[12:59:31] <haggle> https://github.com/joyent/smartos-overlay/blob/master/lib/svc/method/fs-joyent[12:59:43] <defa> haggle: maybe the device names change when messing with the attached drives but this seems strange...[12:59:50] <haggle> that is the boot-file that imports zpools - try running the command manually, perhaps ?[12:59:52] <haggle> device names shouldn't change[13:00:06] <haggle> zfs writes (IIRC) the zpool UUID into the device[13:01:06] <haggle> bah, I meant "device name change shouldn't affect the zpool"[13:01:10] <defa> haggle: I get the prompt for SmartOS Setup … would you like to continue with configuration [Y/n], so I'll type n and try the root password from the image...[13:01:21] <haggle> boot with "noimport=true" IIRC[13:01:23] <haggle> that skips everything[13:01:29] * haggle checks his boot config …[13:01:43] <konobi> you might have smartos=true missing[13:02:03] <haggle> konobi: it works without the new disk, though[13:05:04] <robinbowes> Looks like my problem is this: http://comments.gmane.org/gmane.os.smartos.general/79[13:05:23] <defa> since the system is booted from a non durable ram-disk… where is the configuration to mount the /zones stored?[13:05:54] <robinbowes> defa: in the zpool itself[13:06:09] <robinbowes> (I think)[13:08:34] <haggle> see that script I posted[13:08:44] <robinbowes> In the thread I pasted the link to it says: "If you did need to do this, you'd remove the dump device with dumpadm, delete it with zfs destroy, recreate it with zfs create -V<size> zones/dump, and re-add it with dumpadm."[13:09:00] <robinbowes> I can't figure out how to remove a dump device?[13:12:10] <defa> haggle: I'm having a look at the script and try to figure out what's going on… thanks![13:14:03] *** trentster has quit IRC[13:14:33] <MerlinDMC> robinbowes, should tht caches for dump be disabled (like for swap) ?[13:14:56] *** trentster has joined #smartos[13:23:50] <haggle> defa: if you've got a shell, just spool import, and see what you get[13:23:57] <haggle> also try format<</dev/null and make sure all your drives are listed[13:24:02] <haggle> sorry[13:24:05] <haggle> format</dev/null[13:24:37] <robinbowes> MerlinDMC: sorry, sure what you mean?[13:24:58] <robinbowes> I've tried this:[13:24:59] <robinbowes> zfs create -V 20gb zones/dump[13:25:09] <robinbowes> dumpadm -d /dev/zvol/rdsk/zones/dump[13:25:23] <robinbowes> dumpadm: cannot use /dev/zvol/rdsk/zones/dump as dump device: Block device required[13:29:03] <haggle> try /dev/zvol/dsk/zones/dump ?[13:29:07] <haggle> as in not /rdsk ?[13:29:51] <robinbowes> dump is not supported on device '/dev/zvol/dsk/zones/dump': 'zones' has multiple top level vdevs[13:30:15] <haggle> ah yeah. you've got your disks striped in your zpool ?[13:30:41] <robinbowes> raidz2[13:31:05] <haggle> what smartos release are you running ?[13:31:20] <robinbowes> One moment - just rebooting[13:31:28] <haggle> https://github.com/joyent/illumos-joyent/commit/e6f3a1444022d4d5a78628379c015c0abc48a543 <-- not sure if this will help ?[13:31:37] <robinbowes> (am remote from the box right now)[13:32:12] <robinbowes> Mmm, might do[13:32:26] <robinbowes> Is that in the latest release?[13:32:54] <haggle> I think so - I've got a custom build and it's in there[13:33:10] <haggle> (and it works on my machines which have pure striped disks)[13:33:40] <robinbowes> OK, waiting for machine to come back then will try a remote upgrade[13:34:46] *** enmand has joined #smartos[13:35:24] <robinbowes> Damn, it didn't come back[13:35:28] <haggle> thats never good.[13:35:34] <haggle> lights out ?[13:35:47] <robinbowes> Must have screwed something futzing around with the dump device[13:35:59] <robinbowes> Nah, it's at home[13:36:05] <robinbowes> Will have to wait until tonight[13:36:11] <robinbowes> Nothing spoiling[13:36:12] <haggle> ah, good, at least not a trip to a DC[13:36:22] <robinbowes> Nope[13:36:39] <robinbowes> Oh wait - maybe it is coming back...[13:37:04] <robinbowes> I was getting No route to host[13:37:11] <robinbowes> Am now getting Connection refused[13:37:13] <haggle> surely thats a local problem ?[13:37:16] <haggle> ssh not started yet ?[13:37:27] <robinbowes> Right[13:37:28] <robinbowes> SmartOS Live Image v0.147+ build: 20120517T192048Z[13:37:30] <haggle> (refused means its alive :)[13:37:30] <robinbowes> I'm in[13:37:31] <haggle> yeay[13:38:52] <robinbowes> zpool status looks like this:[13:38:53] <robinbowes> http://pastie.org/4265558[13:39:09] <robinbowes> The device at the bottom with the error was zones/dump[13:39:28] <haggle> oh, so it's written over your disk to write a dump ?[13:39:47] <robinbowes> I deleted the device[13:39:58] <robinbowes> (it did have the device name)[13:43:06] *** enmand has quit IRC[13:43:06] *** jelmd has quit IRC[13:43:06] *** Teknix has quit IRC[13:43:07] *** jperkin has quit IRC[13:43:07] *** Saskaloon has quit IRC[13:43:07] *** reiddraper has quit IRC[13:43:07] *** brendang has quit IRC[13:44:30] *** enmand has joined #smartos[13:44:30] *** jelmd has joined #smartos[13:44:30] *** Teknix has joined #smartos[13:44:30] *** jperkin has joined #smartos[13:44:30] *** Saskaloon has joined #smartos[13:44:30] *** reiddraper has joined #smartos[13:44:30] *** brendang has joined #smartos[13:44:58] <MerlinDMC> robinbowes, was just a question because swap should be created without ram cache i guess (it should be used if no ram is free so caching swap in ram sounds weird); for swap i found articles with smth like this: zfs create -oprimarycache=none -osecondarycache=none ...[14:12:57] *** max` has quit IRC[14:17:12] *** darjeeling has joined #smartos[14:24:40] <defa> How can I mount a zfs formatted disk? sounding like an idiot, I don't understand the options to pass to mount...[14:24:53] <defa> where do I specify the device?[14:26:19] <defa> The disk /dev/dsk/c0t4d0 has a spool from a previous installation where I'd like to save some files from...[14:34:12] <robinbowes> OK, I'm following this: http://wiki.smartos.org/display/DOC/Remotely+Upgrading+A+USB+Key+Based+Deployment[14:34:49] <robinbowes> one thing I notice is that in my usbkey, the existing platform dir is named PLATFORM[14:35:10] <robinbowes> And on the new image, it's named platform[14:35:38] <robinbowes> OK, nm - I checked the grub conf[14:35:44] <robinbowes> It uses lower case[14:37:14] <MerlinDMC> fat32 is not case sensitive so it should not matter[14:37:25] <robinbowes> OK here goes nothing...[14:40:25] <MerlinDMC> what step goes wrong?[14:41:34] <robinbowes> Hopefully, no step[14:41:43] <robinbowes> I'm rebooting for the first time now[14:42:15] <robinbowes> If it doesn't work I'll sort it out when I get home tonight[14:43:38] <nahamu> defa: youre looking for "zpool import"[14:43:42] <nahamu> *you're[14:44:25] <defa> nahamu: thanks… already found that… my boot problems result in two zpools having the same name… now looking into fixing that...[14:45:01] <nahamu> zpool import can use a number to identify a pool, and you can specify a new name when importing a pool[14:49:35] *** denizr has joined #smartos[14:52:26] <robinbowes> Hey, my remote upgrde worked :)[14:52:59] <robinbowes> SmartOS Live Image v0.147+ build: 20120713T005832Z[14:53:09] <nahamu> robinbowes: nice![14:53:22] <robinbowes> Still got the error with the dump device though[14:53:44] <nahamu> might want to delete and recreate the dump device...[14:53:54] <robinbowes> I did[14:53:57] <defa> nahamu: thanks, will try this later… do you know if it is possible to mount/import a zpool within a specific zone?[14:54:00] <nahamu> oh...[14:54:16] <robinbowes> Looks like the new dump device is working OK[14:54:28] <nahamu> defa: I don't think you can do the import from within a zone, but I could be wrong.[14:54:36] <robinbowes> But, I still get the <0x1> error on the old device[14:54:37] <robinbowes> errors: Permanent errors have been detected in the following files:[14:54:38] <robinbowes> <0xc5>:<0x1>[14:54:41] <nahamu> defa: you can do tricks to make the file systems available to it, though.[14:55:02] *** bens1 has joined #smartos[14:55:30] <robinbowes> Any idea how I can clear that?[15:02:47] <nahamu> robinbowes: I don't know enough about dump devices. I'd ask again in a few hours when the west coast of north america is awake.[15:09:34] <MerlinDMC> defa, you can mount a zfs filesystem inside a zone ... but you have to make the zpool available to the GZ[15:10:44] <MerlinDMC> defa, open up the manpage of vmadm and search for "filesystems:"[15:21:23] *** darjeeling has quit IRC[15:31:45] <robinbowes> nahamu: k, will do[15:31:47] <robinbowes> Thanks[15:36:59] *** CarlosC has joined #smartos[15:54:52] *** denizr has quit IRC[15:55:13] *** denizr has joined #smartos[16:08:31] *** jamesd has joined #smartos[16:09:40] *** neophenix has joined #smartos[16:29:09] *** darjeeling has joined #smartos[16:45:50] *** noahmehl has joined #smartos[16:46:42] *** xinkeT has quit IRC[16:47:04] *** haggle has quit IRC[16:47:58] *** Tekni has joined #smartos[16:48:49] <defa> MerlinDMC: thanks! will look into this...[16:53:09] *** jdavis has joined #smartos[16:58:02] *** kfr has joined #smartos[16:58:21] *** deirdres has joined #smartos[17:08:04] *** deirdres has quit IRC[17:26:56] *** dap has joined #smartos[17:34:18] <kfr> I'm curious what kind of drives people are putting in their smartos boxes. I'm thinking of putting 7200RPM seagates[17:37:46] *** denizr has quit IRC[17:38:16] *** denizr has joined #smartos[17:39:47] <Tekni> kfr: I have 10K RPM 2.5" seagate constellation SAS drives[17:41:47] <kfr> I'm going with 1TB SAS Constellation's[17:41:58] *** relling has quit IRC[17:42:08] <Tekni> yeah, those[17:47:14] <kfr> they are only 7200[17:47:18] <kfr> rpm[17:47:58] <e^ipi> i put whatever's on sale in it[17:48:05] <e^ipi> except the WD greens, they're crap[17:50:38] *** denizr has quit IRC[17:52:07] *** denizr has joined #smartos[17:53:44] *** v12 has joined #smartos[17:55:46] <Tekni> kfr: oh yeah, I forgot they're 7200RPM. I also have the 10K 600GB models[17:59:06] *** dap has quit IRC[17:59:47] *** dap has joined #smartos[18:00:12] *** deirdres has joined #smartos[18:03:53] *** noahmehl has quit IRC[18:09:48] *** deirdres has quit IRC[18:14:07] *** mamash has joined #smartos[18:31:27] <kfr> my kvm vm is stuck on stopping[18:31:38] <kfr> there are no qemu pid's for it[18:32:09] <kfr> using joyent_20120614T184600Z[18:33:15] <kfr> tried halt'ing it and it's still stuck on stopping[19:01:50] *** pijewski has quit IRC[19:02:06] *** pijewski has joined #smartos[19:02:36] *** dap has quit IRC[19:03:04] *** dap has joined #smartos[19:06:56] *** Licenser has joined #smartos[19:15:44] *** jdavis has quit IRC[19:15:53] *** kfr has quit IRC[19:16:02] *** kfr has joined #smartos[19:22:09] <kfr> do I need to convert/qemu-img when mving a vm from xen to kvm?[19:23:08] <rmustacc> Depends on how the disk is formatted with xen.[19:23:13] <rmustacc> If it's using some vmdk or special format, then yes.[19:23:19] <rmustacc> If you just have a raw disk image, then it hsould be no.[19:31:19] *** denizr has left #smartos[19:31:43] *** denizr has joined #smartos[19:49:28] *** noahmehl has joined #smartos[19:52:30] *** noahmehl has quit IRC[20:05:31] <CarlosC> rmustacc your ptc work looks interesting...do you know which kind of workflows will be enhanced by this work?[20:06:11] *** bens1 has quit IRC[20:14:22] <rmustacc> The most obvious case are things that basically do object-like behavior without using actual slabs.[20:14:48] <rmustacc> Basically small size allocations that are done repatively or often.[20:18:14] *** bens1 has joined #smartos[20:22:15] *** jdavis has joined #smartos[20:23:55] *** bens1 has quit IRC[20:31:15] *** kfr has quit IRC[20:32:22] <nahamu> rmustacc: is anyone benchmarking its ability to improve performance in particular workloads?[20:37:09] <rmustacc> Not sure if anyone in particular.[20:43:01] <Licenser> :) evening[20:44:55] *** jamesd has quit IRC[20:45:40] *** jamesd has joined #smartos[20:48:14] *** kfr has joined #smartos[20:58:31] <Licenser> FiFo 0.2.0 is closing in :D[21:03:25] *** denizr has quit IRC[21:04:25] *** denizr has joined #smartos[21:26:09] *** jetlag has quit IRC[21:38:27] *** jmeredith has quit IRC[21:46:16] *** deirdres has joined #smartos[21:48:34] <kfr> in the json file for a vm, what should I use for a openindianna vm?[21:49:03] <kfr> virtio for the disk does not work[21:53:20] *** scooby2 has quit IRC[21:54:16] <e^ipi> ide[21:54:39] <e^ipi> xl0 has some virtio drivers you can slip in after the fact, but they're still very experimental[21:54:49] *** mamash has left #smartos[21:54:53] <richlowe> he's actually got them out for review to get them integrated again[21:56:04] *** deirdres has quit IRC[21:56:49] <kfr> what do I specify for the nic is I don't want to use virtio?[21:57:20] *** deirdres has joined #smartos[21:57:56] <e^ipi> e1000 works[21:58:38] <e^ipi> richlowe: oh, has he? not gonna lie, haven't been paying enough attention to -discuss[21:58:41] <e^ipi> err, -developer[21:59:02] <e^ipi> oh yeah, there it is[22:01:31] *** noahmehl has joined #smartos[22:02:40] *** scooby2 has joined #smartos[22:17:08] *** jmeredith has joined #smartos[22:19:04] *** enmand has quit IRC[22:24:35] *** noahmehl_ has joined #smartos[22:24:35] *** noahmehl has quit IRC[22:24:35] *** noahmehl_ is now known as noahmehl[22:45:29] *** bens1 has joined #smartos[22:51:30] *** neophenix has quit IRC[22:57:10] *** jamesd has quit IRC[22:57:16] *** jamesd_ has joined #smartos[22:59:15] *** haggle has joined #smartos[22:59:48] *** marsell has joined #smartos[23:00:13] <haggle> rmustacc: Re. the Perl funny stuff yesterday - I rebuilt on a fresh box and all is working. Sorry for the hassle yesterday :)[23:02:28] *** bens1 has quit IRC[23:03:06] <rmustacc> No worries, glad it's sorted out.[23:04:29] *** defa has quit IRC[23:05:45] *** noahmehl_ has joined #smartos[23:05:46] *** noahmehl has quit IRC[23:05:46] *** noahmehl_ is now known as noahmehl[23:06:54] *** dap has quit IRC[23:09:30] *** bixu has quit IRC[23:09:41] *** banjiewen|away has quit IRC[23:09:50] *** dap has joined #smartos[23:09:58] *** bixu has joined #smartos[23:10:15] *** deirdres has quit IRC[23:19:23] *** banjiewen|away has joined #smartos[23:20:33] *** CarlosC has quit IRC[23:35:50] <_jared_> anyone around who can talk SMF?[23:36:59] <rmustacc> I know a little bit.[23:37:02] <rmustacc> What's up?[23:37:26] <_jared_> any way to specify a process should be ignored in a SMF manifest[23:37:48] <_jared_> we have something that starts with svcadm and it has child processes that we don't ever want to shut down[23:37:59] *** deirdres has joined #smartos[23:38:07] <_jared_> right now, svcadm disable <app> just puts the app in maintenance mode[23:38:44] <rmustacc> Okay, so the behavior of smf depends on the contract type that you have.[23:38:47] <rmustacc> So you have a few options.[23:39:04] <rmustacc> Put another way, when do you want the service to be restarted.[23:39:17] <rmustacc> When all the programs that are launched exit, when only a single one exits, something else?[23:39:56] <_jared_> the top level app can be started, restarted, refreshed at will, but the child process should never be killed[23:40:32] <rmustacc> What should happen when the child process restarts?[23:40:42] <rmustacc> Or when it dies.[23:41:36] <_jared_> it would only die in exceptional circumstances[23:41:46] <_jared_> but if it did, it'd be nice to restart it[23:42:19] <rmustacc> Right, so the semantics of that are important.[23:42:23] <_jared_> if it helps, it is epmd (erlang port mapper daemon)[23:42:52] <rmustacc> It sounds like you want two different services.[23:42:55] <_jared_> but we aren't using the packaged erlang, so we can't use the semantics of svc:network/epmd[23:43:30] <rmustacc> Why not use that same basic one, just change the name and start method to point to yours?[23:43:32] <_jared_> admittedly we aren't playing very nice, that's why I'm trying to figure out a way that doesn't make other erlang apps mad[23:43:46] <_jared_> well we ship erlang in our app[23:44:06] <rmustacc> You can ship more than one manifest and have your app depend on your version of epmd.[23:44:06] <_jared_> hold on let me gather some links[23:45:16] <_jared_> pfff, damn github, if it wasn't saying your PK repo was down, I could give you some context[23:45:50] <_jared_> so rabbit, couch all have the notion of @SMF_NAME@-epmd like you mentioned[23:46:20] <_jared_> but there is still just one actual "epmd" daemon according to 'ps -ef'[23:46:54] <_jared_> so i was hoping to have our app start epmd if it isn't already started, but not kill it or depend on it when our app shuts down[23:47:18] <_jared_> i looked in the SMF docs for something I can put in the manifest file, but didn't see anything appropriate[23:47:21] <rmustacc> Sounds to me like your app should just depend on the epmd daemon right?[23:47:39] <rmustacc> Usually that should cause it to start dependent services, iirc.[23:48:03] <_jared_> yeah, but "which" epmd is the question, because the ones in your pk repo all point to say /opt/local/erlang/bin/epmd[23:48:10] <_jared_> but ours would be in our app directory[23:48:25] <rmustacc> Then you need to create your own service.[23:48:50] <_jared_> so is having a different "name=" field sufficient?[23:48:55] <rmustacc> And your start method could always basically if epmd is running, do nothing. If it's not, do something different.[23:48:59] <rmustacc> Oh, yeah.[23:49:01] *** denizr has quit IRC[23:49:02] <_jared_> even if the process name still shows as "epmd"[23:49:05] <rmustacc> Yup[23:49:09] <_jared_> oh ok[23:49:14] <rmustacc> You can have multiple instances of anything running.[23:49:21] <rmustacc> The management is done by contract, it doesn't care what the execname is.[23:49:24] *** denizr has joined #smartos[23:49:28] <_jared_> ok[23:49:36] <_jared_> well we'll see what epmd thinks about that[23:49:37] <rmustacc> A common example is having multiple instances of apache.[23:49:55] <rmustacc> Yeah, if epmd doesn't like having two copies running, then you have to figure out something else.[23:50:12] <_jared_> have my install script uninstall rabbit and couch[23:50:14] <_jared_> :)[23:50:32] <rmustacc> If you're customers are fine with that, sounds like a plan.[23:50:36] <rmustacc> *your[23:50:46] <_jared_> I doubt it, but I'll figure something out[23:50:48] <_jared_> thanks[23:51:33] <_jared_> so I take it you can depend on a service and tell it to start it up if it isn't already started, but to never kill it when you shut down?[23:52:13] <_jared_> nm, that's a dumb question, because we depend on the filesystem service and that certainly doesn't stop on shut down[23:52:14] <rmustacc> If you depend on a service, by shutting down you should not stop a dependency.[23:52:30] <rmustacc> If you stop a dependency, then you should be stopped as well.[23:52:31] <_jared_> ok, thanks rmustacc[23:53:13] <rmustacc> Sure, no problem.[23:53:47] *** PiotrSikora has quit IRC[23:56:07] *** PiotrSikora has joined #smartos[23:56:58] *** deirdres has quit IRC