September 25, 2013 
< | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | >
September 25, 2013 [00:01:26] *** neophenix has quit IRC[00:06:07] *** julianduque has quit IRC[00:08:14] <lpsmith> oddsignals, of course at the moment GHC running in Linux KVM might actually be better, depending on what you are doing[00:08:51] <lpsmith> There is no event port support in GHC's IO manager, so it defaults back to the select (or poll?) implementation[00:09:08] <lpsmith> That's what really started this, I was interested in trying to port GHC's IO manager to event ports.[00:09:56] *** sparq_ has joined #smartos[00:09:57] <lpsmith> So for certain server applications, GHC is best used on Linux or FreeBSD or MacOSX[00:10:57] <lpsmith> But the worst problems on that count shouldn't be _that_ hard to fix.[00:11:58] *** sparq_ has quit IRC[00:20:36] <oddsignals> lpsmith: well, it's certainly the least time-consuming option at the moment :-)[00:22:35] <oddsignals> lpsmith: good luck in any case![00:27:06] *** elijah-mbp has quit IRC[00:29:08] *** elijah-mbp has joined #smartos[00:29:37] *** Fuzai has quit IRC[00:30:48] *** julianduque has joined #smartos[00:30:54] *** julianduque has joined #smartos[00:38:34] *** Fuzai has joined #smartos[00:40:06] *** Fuzai has quit IRC[00:40:33] *** mary5030 has quit IRC[00:42:26] *** Fuzai has joined #smartos[00:44:39] *** W0rmDr1nk has quit IRC[00:44:51] *** Fuzai has quit IRC[00:45:25] *** Fuzai has joined #smartos[00:46:31] *** Fuzai has quit IRC[00:48:44] *** Fuzai has joined #smartos[00:51:45] *** Fuzai has quit IRC[00:56:29] *** axonpoet has quit IRC[00:58:18] *** jim80net has quit IRC[01:09:08] *** jsm has quit IRC[01:10:52] *** lloydde has quit IRC[01:11:27] *** lloydde has joined #smartos[01:15:33] *** lloydde has quit IRC[01:40:40] *** lloydde has joined #smartos[01:47:38] *** badboy_ has quit IRC[01:47:39] *** hij1nx_onaboat has quit IRC[01:47:43] *** Hisa_X has quit IRC[01:47:44] *** ed2091 has quit IRC[01:47:48] *** Azbruh has quit IRC[01:47:49] *** copec has quit IRC[01:47:51] *** pinja has quit IRC[01:47:53] *** oc has quit IRC[01:47:57] *** jesse_ has quit IRC[01:47:59] *** mburns has quit IRC[01:47:59] *** while1eq1 has quit IRC[01:48:07] *** Alasdairrr has quit IRC[01:48:13] *** theCzar has quit IRC[01:48:15] *** sebasp_ has quit IRC[01:48:19] *** potatosalad has quit IRC[01:48:28] *** sa5bke has quit IRC[01:48:35] *** Niamkik_ has quit IRC[01:48:39] *** aszeszo has quit IRC[01:48:42] *** jones_d has quit IRC[01:48:43] *** trentster has quit IRC[01:48:45] *** djun has quit IRC[01:48:48] *** elijah-mbp has quit IRC[01:49:00] *** nshalman has quit IRC[01:49:14] *** halldorh has quit IRC[01:49:15] *** enmand has quit IRC[01:49:16] *** dlg has quit IRC[01:49:20] *** e^ipi has quit IRC[01:49:21] *** Licenser has quit IRC[01:49:22] *** scanf has quit IRC[01:49:22] *** rmustacc has quit IRC[01:49:24] *** ashley_w_ has quit IRC[01:49:27] *** bixu has quit IRC[01:49:31] *** dap has quit IRC[01:49:31] *** papertigers has quit IRC[01:49:35] *** defel has quit IRC[01:49:36] *** Raynos has quit IRC[01:49:39] *** pde has quit IRC[01:49:41] *** |woody| has quit IRC[01:49:41] *** richlowe has quit IRC[01:49:49] *** mmalecki has quit IRC[01:49:55] *** saltsa has quit IRC[01:49:57] *** arekinath has quit IRC[01:49:59] *** russiane39 has quit IRC[01:50:04] *** ImTheBitch has quit IRC[01:50:05] *** Xenith has quit IRC[01:50:05] *** cmang_ has quit IRC[01:50:07] *** njm has quit IRC[01:50:10] *** pettson has quit IRC[01:50:11] *** Alasdairrr has joined #smartos[01:50:11] *** jesse_ has joined #smartos[01:50:16] *** Azbruh has joined #smartos[01:50:24] *** mburns has joined #smartos[01:55:16] *** pettson has joined #smartos[01:55:16] *** njm has joined #smartos[01:55:17] *** cmang_ has joined #smartos[01:55:17] *** Xenith has joined #smartos[01:55:17] *** ImTheBitch has joined #smartos[01:55:17] *** russiane39 has joined #smartos[01:55:17] *** arekinath has joined #smartos[01:55:17] *** saltsa has joined #smartos[01:55:17] *** oc has joined #smartos[01:55:17] *** copec has joined #smartos[01:55:17] *** Hisa_X has joined #smartos[01:55:17] *** badboy_ has joined #smartos[01:55:17] *** hij1nx_onaboat has joined #smartos[01:55:17] *** while1eq1 has joined #smartos[01:55:18] *** 64MAAO4XW has joined #smartos[01:55:18] *** elijah-mbp has joined #smartos[01:55:18] *** bixu has joined #smartos[01:55:18] *** theCzar has joined #smartos[01:55:18] *** sebasp_ has joined #smartos[01:55:18] *** dap has joined #smartos[01:55:18] *** pde has joined #smartos[01:55:18] *** potatosalad has joined #smartos[01:55:18] *** nshalman has joined #smartos[01:55:18] *** sa5bke has joined #smartos[01:55:18] *** papertigers has joined #smartos[01:55:18] *** halldorh has joined #smartos[01:55:19] *** defel has joined #smartos[01:55:19] *** enmand has joined #smartos[01:55:19] *** dlg has joined #smartos[01:55:19] *** Niamkik_ has joined #smartos[01:55:19] *** Raynos has joined #smartos[01:55:19] *** aszeszo has joined #smartos[01:55:19] *** Licenser has joined #smartos[01:55:19] *** e^ipi has joined #smartos[01:55:19] *** jones_d has joined #smartos[01:55:19] *** trentster has joined #smartos[01:55:19] *** ashley_w_ has joined #smartos[01:55:19] *** djun has joined #smartos[01:55:19] *** |woody| has joined #smartos[01:55:20] *** mmalecki has joined #smartos[01:55:20] *** richlowe has joined #smartos[01:55:20] *** rmustacc has joined #smartos[01:55:20] *** scanf has joined #smartos[01:55:36] *** lloydde has quit IRC[01:55:38] *** lloydde has joined #smartos[01:56:13] *** lloydde has quit IRC[02:00:38] *** theCzar has quit IRC[02:03:53] *** ryancnelson has quit IRC[02:08:10] *** pinja has joined #smartos[02:16:29] *** theCzar has joined #smartos[02:17:32] *** jsm has joined #smartos[02:18:25] *** theCzar has quit IRC[02:20:59] *** julianduque has quit IRC[02:23:59] *** julianduque has joined #smartos[02:32:03] *** mary5030 has joined #smartos[02:34:12] *** mary5030 has quit IRC[02:34:48] *** mary5030 has joined #smartos[02:38:56] *** axonpoet has joined #smartos[02:50:08] *** lloydde has joined #smartos[02:57:04] *** lloydde has quit IRC[02:57:31] *** lloydde has joined #smartos[03:00:20] *** lloydde_ has joined #smartos[03:00:24] *** lloydde_ has quit IRC[03:00:39] *** lloydde has quit IRC[03:00:50] *** lloydde has joined #smartos[03:05:08] *** lloydde has quit IRC[03:19:41] *** potatosalad has quit IRC[04:03:08] *** Thrae has quit IRC[04:03:09] *** wolstena has joined #smartos[04:10:11] *** Thrae has joined #smartos[04:11:26] *** lloydde has joined #smartos[04:16:37] *** lloydde has quit IRC[04:18:56] *** Thrae has quit IRC[04:23:12] *** Thrae has joined #smartos[04:35:17] *** ryancnelson has joined #smartos[04:39:56] *** leroux has joined #smartos[04:55:11] *** lloydde has joined #smartos[05:01:23] *** lloydde has quit IRC[05:01:49] *** lloydde has joined #smartos[05:06:42] *** lloydde has quit IRC[05:16:50] *** wolstena has quit IRC[05:17:15] *** ipalreadytaken has joined #smartos[06:02:53] *** ipalreadytaken has quit IRC[06:05:03] *** potatosalad has joined #smartos[06:06:03] *** julianduque has quit IRC[06:06:27] *** szaydel has quit IRC[06:07:21] *** lloydde has joined #smartos[06:26:16] *** LnxTx has quit IRC[06:26:31] *** LnxTx has joined #smartos[06:26:45] *** bahamas10 has quit IRC[06:26:55] *** defel has quit IRC[06:27:30] *** defel has joined #smartos[06:27:55] *** bahamas10 has joined #smartos[06:30:04] *** nefilim has quit IRC[06:34:06] <opeth__> A[06:34:14] <opeth__> damn, not here[06:39:27] *** axonpoet has quit IRC[06:49:55] *** lloydde has quit IRC[06:50:22] *** lloydde has joined #smartos[06:52:33] *** lloydde_ has joined #smartos[06:52:39] *** lloydde has quit IRC[06:57:23] *** mary5030 has quit IRC[07:24:00] *** lloydde_ has quit IRC[07:24:26] *** lloydde has joined #smartos[07:28:56] *** lloydde has quit IRC[08:04:04] *** opeth__ has quit IRC[08:09:39] *** mamash has joined #smartos[08:14:15] *** noahmehl has quit IRC[08:14:21] *** noahmehl_ has joined #smartos[08:15:36] *** axonpoet has joined #smartos[08:25:33] *** lloydde has joined #smartos[08:29:43] *** lloydde has quit IRC[08:31:11] *** hadret has joined #smartos[08:46:03] *** axonpoet has quit IRC[08:59:08] *** tonist has joined #smartos[09:00:29] *** hotbox has joined #smartos[09:03:19] *** yruss972 has joined #smartos[09:15:48] *** CaTtleyA has joined #smartos[09:20:01] *** texarcana has quit IRC[09:22:06] *** texarcana has joined #smartos[09:24:13] *** ryancnelson has quit IRC[09:25:55] *** lloydde has joined #smartos[09:29:20] *** tonist has joined #smartos[09:30:22] *** lloydde has quit IRC[09:35:01] *** jamesog has quit IRC[09:36:40] *** jamesog has joined #smartos[09:42:25] *** yruss972 has quit IRC[09:43:01] *** yruss972 has joined #smartos[09:47:16] *** yruss972 has quit IRC[09:48:29] *** alcir has joined #smartos[09:50:25] <oddsignals> Anyone seen this before? vmadm info {uuid} -> Unable to get VM info for VM {uuid}: connect ECONNREFUSED[09:50:31] <oddsignals> same went for vmadm stop[09:50:45] <oddsignals> reboot solved it, just wondering what might have triggered it[09:51:04] <konobi> did you check `svcs -x` ?[09:52:22] <oddsignals> I did not :-) will do if it turns up again[09:55:07] <oddsignals> I think whatever caused it might have been triggered by creating a new KVM VM with the same VNC port as another VM[09:57:23] *** bens1 has joined #smartos[09:58:09] *** KermitTheFragger has joined #smartos[10:04:54] *** phyrne has joined #smartos[10:08:34] <phyrne> Hey guys, I'm new here. I'm looking at trying out SmartOS on a little home server I've got, been reading up on it and it seems absolutely awesome. I'm currently running OpenIndiana, kind of looking for a "better solution" due to some aspects I haven't been particularly comfortable with. I have a question I hope some of you may be able to answer. The zpool I have right now was created with ZFS v5, does SmartOS run ZFS 5 (or later), as this would make the trans[10:11:18] *** yruss972 has joined #smartos[10:12:37] *** leecallen35 has joined #smartos[10:15:55] *** yruss972 has quit IRC[10:16:27] *** leecallen has quit IRC[10:20:21] <MerlinDMC> oddsignals, that happens when the vmadmd service is not running anymore ... and a reboot did probably restart that service for you[10:23:18] <MerlinDMC> phyrne, it's zpool v28 (+ feature flags) and zfs v5 for the filesystems on that pool[10:23:42] <phyrne> MerlinDMC: Excellent, thank you :)[10:26:15] *** lloydde has joined #smartos[10:30:14] *** lloydde has quit IRC[10:44:26] *** andy_js has joined #smartos[10:52:31] *** W0rmDr1nk has joined #smartos[11:12:45] *** opeth__ has joined #smartos[11:13:49] *** yruss972 has joined #smartos[11:17:34] *** yruss972 has quit IRC[11:18:09] *** yruss972 has joined #smartos[11:20:59] *** CaTtleyA has quit IRC[11:22:39] *** yruss972 has quit IRC[11:22:42] *** CaTtleyA has joined #smartos[11:26:36] *** lloydde has joined #smartos[11:31:33] *** lloydde has quit IRC[12:26:19] *** tonist has quit IRC[12:28:41] *** yruss972 has joined #smartos[12:33:02] *** yruss972 has quit IRC[12:45:02] <phyrne> Would it be sensible to use SmartOS for a little home server I'm running, who's primary purpose is to serve up files (zfs and nfs)? I will also be running some VMs. So far, from what I've read, SmartOS is an excellent solution for someone looking to run an illumos based distro. I'd just like some of your opinions on using it primarily geared towards using it as a fileserver (backups and media)[12:48:49] <jesse_> nfs doesn't work in zones and setting it up to the global zone is a pain (afaik)[12:49:20] <jesse_> the native cifs support is the same, but you can run samba in a zone[12:49:45] <phyrne> Hm, okay, thanks jesse_. I did have the idea in my head that I'd be using the global zone to do a lot of the work I was looking for[12:49:54] <arekinath> phyrne: probably though omnios is a better choice for something that's only doing file serving[12:50:05] <arekinath> since it has a global zone you can install stuff in and keep config like nfs/smb stuff[12:51:58] <phyrne> arekinath: Great, thank you. I'm still very interested in SmartOS, but perhaps for this particular purpose, I should look for a more persistent configuration, taking into account the fact that /etc is overwritten on each boot and other such things. Not exactly geared towards what I was looking for in this situation. Seems like an awesome solution for virtualization though![12:52:57] <arekinath> phyrne: yeah, if you expect to do lots of vms on it, smartos is probably better... if it's primarily doing file serving and a few ancillary tasks, then one of the other distros is likely easier to work with[12:53:00] <jperkin> right, SmartOS is heavily focused on virtualisation, you can do other things but they will always be somewhat hacky compared to other distributions[12:53:12] <jperkin> OmniOS is a fine choice, and you can still use our packages on it[12:55:06] <arekinath> omnios is stealing the kvm brand soon actually. or it's on their roadmap at least. :P[12:55:22] <arekinath> not sure how it'll be configured though[12:56:11] <oddsignals> phyrne: I'm using an OmniOS box as a fileserver and a SmartOS for VMs[12:57:02] <phyrne> Great, thanks guys. This all pretty much confirms my thoughts :)[12:57:46] <jperkin> I thought OmniOS had all the KVM stuff for ages, it was just the tooling they didn't have (because they don't want to depend on node)[12:58:21] <arekinath> jperkin: yeah, kvm runs of course, but setting up kvm in a zone is a bit of a pain at the moment[12:59:09] <phyrne> I currently have OpenIndiana running on the server (only a few days), but there are some inconsistencies that have made me a bit uncomfortable, and makes me question the development model they're following. I love the sound of Omni, thanks people! Also, SmartOS will absolutely be getting implemented in my home for a little cluster, I was thinking of going with something like proxmox a while back, but this seems like a much better, less-linuxy (:p) solution[12:59:09] <arekinath> jperkin: all the instructions say to just run your kvms in the gz. it's nicer to put it in a zone though[12:59:56] <arekinath> phyrne: yeah, openvz is like a pale buggy imitation of solaris zones. ;)[13:00:09] <phyrne> Haha[13:00:40] <jperkin> ok[13:01:04] <phyrne> I'm a noob when it comes to illumos (and Solaris), I'm a BSD user, but I'm loving this so far, it's a whole different world[13:01:44] *** AlainODea has left #smartos[13:01:53] <jperkin> at least you'll be used to the packaging then[13:02:19] <phyrne> Yeah, I've used pkgsrc on NetBSD quite a bit[13:02:31] <phyrne> Really enjoying digging into ZFS too[13:03:19] <phyrne> So, I imagine I'll frequent this channel fairly regularly in the future, you folks sure are helpful. I popped into OpenIndiana the other day, and nobody could answer my questions :/[13:05:55] *** tonist has joined #smartos[13:07:17] *** mamash has left #smartos[13:07:27] *** mamash has joined #smartos[13:27:18] *** lloydde has joined #smartos[13:30:25] *** szaydel has joined #smartos[13:31:39] *** lloydde has quit IRC[13:35:17] *** tonist has quit IRC[13:36:25] *** tonist has joined #smartos[14:06:07] *** CaTtleyA has quit IRC[14:07:01] *** CaTtleyA has joined #smartos[14:16:07] *** joshie has quit IRC[14:22:00] *** khushildep has joined #smartos[14:22:41] <khushildep> Hey all[14:22:42] <khushildep> Quick question, after boot, is there any reason why I have to have the USB key/ISO mounted all the time?[14:24:23] <arekinath> khushildep: nope, once everything's in ram you can take it away as you like[14:24:37] <arekinath> khushildep: but if it reboots it's good to have something to boot off of ;)[14:25:51] <khushildep> nice - thanks[14:27:39] *** lloydde has joined #smartos[14:32:02] *** lloydde has quit IRC[14:38:22] *** andy_js has quit IRC[14:39:52] *** andy_js has joined #smartos[14:54:07] *** mikl has joined #smartos[15:01:52] <leecallen35> pbyrne: are you still here?[15:09:07] <opeth__> jperkin: would you have any ideas why postgrey fails to daemonize whenever the taint checks are enabled? this isn't pkgsrc-specific and isn't even terribly disquieting but it might call for a patch to disable that as hard-coded in the packaged script... unless solved[15:10:32] <opeth__> OTOH, I've migrated the mail stack to a 13.2 vm on the locally rolled pkgsrc packages with versions and supports I needed and all is green... but I've had to remove -T from postgrey the very way I had to earlier[15:11:54] <jperkin> dunno, I briefly used it years ago but didn't really like it[15:12:01] <jperkin> it doesn't seem to be well maintained[15:12:41] <opeth__> I supposed this has something to do with the way certain perl modules down the line might have been compiled[15:12:53] <opeth__> I wish I remembered if that surfaced on S11 too[15:13:03] <jperkin> I last used gld which had some child handling issues, before giving up with greylisting[15:13:09] <opeth__> it does on SmartOS for sure, but I can live with it[15:13:37] <jperkin> you could always try running it with /usr/perl5/bin/perl to rule out pkgsrc-specific perl issues[15:13:44] <opeth__> I think I'll generate a diff and add it to my package for the future[15:14:01] <opeth__> as a patch I mean[15:14:09] <opeth__> oh, that's a good idea[15:15:14] <opeth__> I only aim to make this as hack-free and automation-enabled as possible now that I've managed to ditch Gentoo Prefix on SmartOS :][15:20:51] <mamash> i'm pretty sure there were some pkgsrc patches committed to mail/postgrey after 2013Q2 was cut[15:22:06] <mamash> related to taint checks[15:27:06] <mamash> http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/mail/postgrey/patches/patch-postgrey?rev=1.2[15:27:17] *** jim80net has joined #smartos[15:28:00] *** lloydde has joined #smartos[15:28:25] <jperkin> Successfully built: 11017[15:28:28] <jperkin> nice, we hit 11k[15:32:30] *** lloydde has quit IRC[15:37:00] <Alasdairrr> :-)[15:37:37] <phyrne> leecallen35: Sorry, I was out to lunch, I'm here now. Sup? :)[15:38:29] *** CaTtleyA has quit IRC[15:38:31] <leecallen35> the SmartOS as NAS is kind of a FAQ. and the answers given are usually, no, it doesn't work.[15:38:34] <leecallen35> but it does.[15:39:01] <leecallen35> it has some imperfections, those were pointed out earlier, including the fact NFS is limited to the GZ[15:39:10] <leecallen35> but here is the workaround some of us use...[15:39:17] <phyrne> leecallen35: Thanks, yeah I got some great feedback earlier, so, I believe I'm going with OmniOS[15:39:33] <phyrne> After reading up on it, it too seems excellent[15:39:42] <leecallen35> run Samba in a zone. share your GZ ZFS datasete with the Samba zone via lofs.[15:39:55] <leecallen35> now you can share your data with NFS from the GZ, and using SMB in a zone.[15:40:15] <leecallen35> yes I would recommend OmniOS too -- you can use Napp-it which is really cool - unless you also need virtualization[15:40:20] <jperkin> it's not that it doesn't work, it's that if file serving is your primary aim then SmartOS is probably the worst choice you could make ;)[15:40:44] <leecallen35> I would not recommend it unless you also need virtualization[15:40:45] <phyrne> jperkin: Haha, yes, I cam to realise this after reading some and then speaking to you guys :p[15:41:15] <jperkin> but that's one of the things I love about SmartOS, the focus on doing one thing well, and being happy to say it sucks at other stuff.[15:41:20] <phyrne> I certainly don't need virtualization, however, I will have a dedicated host in my network which I'll put SmartOS on for a mini cluster :)[15:41:29] <phyrne> jperkin: Absolutely[15:41:32] <leecallen35> good[15:41:33] <jperkin> too many OS try to be mediocre at everything, and end up that way[15:41:45] <Alasdairrr> We've used smartos as an nfs server very successfully, because the share data is stored as zfs properties it works great upon reboots :-) COMSTAR however is less successful as the state disappears on reboot...[15:41:50] <phyrne> ^ Linux :p (Not bashing, I use it at work)[15:42:30] <leecallen35> I agree about that feature of ZFS/NFS - it avoids the persistence problem (/etc/exports)[15:42:52] <leecallen35> anyway I am drafting a FAQ for this, it comes up so often, I will ask the community here to mark it up for me[15:42:58] <phyrne> Alasdairrr: Yeah, this is a little microserver for home, I'll be using it for sharing media, doing baskups to it, and perhaps some other tasks. Omni seems best suited :)[15:43:13] <phyrne> leecallen35: Great, that's a good idea :)[15:43:20] <Alasdairrr> Yup, we're a datacenter environment with PXE netboot servers so SmartOS ends up just about everywhere here[15:43:47] <Alasdairrr> plug in server, enable PXE on the network cards, boom, smartos[15:43:56] <Alasdairrr> :-)[15:44:06] <leecallen35> slick[15:45:22] <phyrne> Alasdairrr: Awesome. I do have a question for you actually, if you don't mind me asking; is the server build of OI built from the desktop build, just with "stuff" taken out? I'm curious as OI's filesystem seems to have some artifacts (some files/directories) relating to software that is absent, for instanc: startx, /etc/X11 and even /usr/share/gnome[15:46:28] <Alasdairrr> The last step is making the ISOs, the distro-constructor software just pkg installs particular sets of packages to a proto area[15:46:37] <Alasdairrr> so the server build is one set of packages, the desktop build is another[15:47:08] <Alasdairrr> the artifacts are due to directories/files living in particular packages, like SUNWcs[15:47:23] <Alasdairrr> There's a lot of history/heritage in the way its all done[15:47:45] <phyrne> Ah I see :) Okay, thanks for explaining, I was curious[15:47:47] <Alasdairrr> np[15:48:12] <Alasdairrr> Feel free to file issues on the bug tracker for stuff like that, and if you feel so inclined, feel free to download the source and submit patches[15:48:28] <phyrne> Great, will do![15:48:43] <Alasdairrr> although thats easier said than done due to the way Sun assembled the whole OS, there is no single source repo to download, the OS is fragmented into lots of bits that were built by different teams[15:49:13] <Alasdairrr> each bit being a "consolidation", such as the JDS desktop stuff (Gnome + random crap), SFW, userland-gate, etc[15:50:57] <phyrne> Yup, certainly makes for an interesting structure. When reading through some of the OmniOS docs earlier, I thought it was great that software was located in specific repos in accordance with level of necessity/purpose[15:51:29] <phyrne> I do apologise if my terminology is a little off, I'm new to Solaris/illumos[15:51:58] *** chorrell has quit IRC[15:52:25] <Alasdairrr> no mistakes observed so far ;-)[15:53:31] <phyrne> Hah, well, I'm glad I seem to be grasping this. It's quite some time since I've been so "excited", so to speak, to discover a new family of operating systems[15:55:04] <phyrne> I've always heard of Solaris of course, but never thought much to look into it. It's really ZFS that lead me to find out about illumos[15:55:28] *** CaTtleyA has joined #smartos[15:56:34] <Alasdairrr> Yeah ZFS is awesome :-)[15:59:06] <phyrne> Indeed! I'm loving it[16:01:17] *** neophenix has joined #smartos[16:02:09] <phyrne> The only volume manager I'd worked with before was LVM at work, other than that, just FFS2 at home :p I do wish OpenBSD would get ZFS support. Mind you, that said; I'm enjoying learning Solaris systems[16:04:44] <phyrne> Alasdairrr: You're from around London aren't you? (Yes I watched the OI talk on YouTube) Are there any Solaris/illumos events/meetups that go on around here?[16:05:31] *** hadret has quit IRC[16:05:40] <Alasdairrr> Yeah, I live in Greenwich at the moment and work at London Bridge. There are the Oracle UKOUG Solaris meetups which I stopped going to for obvious reasons but there was a guy trying to start some illumos specific ones[16:06:24] <Alasdairrr> http://www.meetup.com/solarians/[16:08:00] <phyrne> Alasdairrr: London Bridge? I'm on Fenchurch Street :) Ah, I see, fair enough. Cheers! I've never been to any *nix oriented meetups... or any meetups for that matter[16:08:49] <jperkin> if people could occasionally broadcast meetups on here it'd be useful, I'd quite like to make one, but I don't want to sign up for yet more services.[16:08:59] <Alasdairrr> I used to go to the Sun Solaris ones, LOSUG, the London OpenSolaris Users Group[16:09:06] <phyrne> jperkin: You and me both :p[16:13:16] *** chrispercol has joined #smartos[16:14:22] <chrispercol> hi, how do I manage private keys on smartos? I can see an option for authorized keys, is there something similar for private keys? I have setup /opt as a git repo and want to be able to push it with my key[16:15:13] <jperkin> no, you'd need to do it manually, or just do what I do and 'ssh -A' if I want to push something from the GZ[16:15:22] <MerlinDMC> chrispercol, agent forwarding not an option?[16:16:26] <phyrne> Has anyone used Puppet on OmniOS?[16:16:59] *** johnnuttall has joined #smartos[16:18:04] <chrispercol> thanks guys, I'll go from the other direction and pull the repo rather than push it and only use authorized keys then[16:18:34] <chrispercol> I thought putting private keys on the usb was prob a bad idea anyway[16:18:37] *** wolfeida_ has joined #smartos[16:20:53] *** wolfeidau has quit IRC[16:23:00] *** chrispercol has quit IRC[16:28:20] *** lloydde has joined #smartos[16:29:20] *** wolfeidau has joined #smartos[16:29:28] <Jared_M> I have kvm image I'm trying to clone and copy to another smartos machine, but zfs claims its filesystem is taking up 60gigs of space when it is only taking up 5.7gigs in reality. I tried the typical dd if=/dev/zero trick on the Linux vm, but that made it jump from 32gig to its current 65gig of space according to zfs list. Knowing that zfs is append-only, i see why this happened, but is there a way to tell zfs to garbage[16:29:28] <Jared_M> collect just that one zvol?[16:29:45] <Alasdairrr> Does OpenVPN server work inside a SmartOS Zone?[16:29:48] *** wolfeida_ has quit IRC[16:30:52] <yofuh> Alasdairrr: yes, works fine (at least with tap devices)[16:31:15] <Alasdairrr> Cool, thanks yofuh[16:32:20] <jesse_> Jared_M, not a 'proper' solution, you could use zfs compression to get all the zeroed-out pages to take less space. And maybe remove the compression after the transfer (and before using the image again)[16:32:38] *** lloydde has quit IRC[16:32:42] *** tonist has quit IRC[16:34:03] <Jared_M> when dumping the image and piping through gzip, I'm sure that will make the image smaller, but just wanted to clean out the cruft before cloning this thing X times[16:34:18] <Jared_M> what does joyent do to prep images for imgadm?[16:34:45] <jesse_> I mean enable zfs compression, and write the blocks again with zeroes (I assume that's what you did in linux)[16:35:39] <jesse_> afaik, that should compress the blocks with all-zeroes, taking less space than written out[16:35:51] <Jared_M> I can try that[16:35:54] <MerlinDMC> Jared_M, https://github.com/joyent/smartos-vmtools ... there is a outdated but functional version of a prep script for linux[16:35:55] <jesse_> but it's not the same as not-having-allocated-the-blocks-to-begin-with[16:35:56] *** jim80net has quit IRC[16:36:04] <Jared_M> yes[16:36:26] *** johnnuttall has quit IRC[16:36:37] *** jim80net has joined #smartos[16:36:56] <jesse_> one would be to make another disk for the linux, and then do mkfs/grub/rsync to that from the boot disk[16:37:07] <MerlinDMC> after the prep it's zfs snapshot + zfs send piped through gzip/bzip2[16:37:11] <jesse_> I guess that would only allocate what it needs[16:37:40] <jesse_> MerlinDMC, the problem here seems to be that the image is too big before snapshot[16:38:25] <Jared_M> well if the resulting image that I need to transfer between machines is small, I suppose that's all that is really important[16:38:49] <Jared_M> MerlinDMC: thanks for the link to the vmtools, I'll take a look[16:40:11] <MerlinDMC> Jared_M, if it's a one time thing and you want to save traffic ... pipe it through pbzip2 and don't care about the nulled blocks - bzip2 decompression will be slow tho[16:41:52] <nahamu> is there a userspace lz4 compressor?[16:42:28] <MerlinDMC> nahamu, yes ... but compression ratio is low ;) xz would be fine as well ... has decent comression and high decompression speed[16:42:37] <nahamu> would be nice to be able to pipe through an lz4 compressor, fast, but decent compression.[16:44:07] <leecallen35> I have created a draft wiki entry to answer the frequently asked question about SmartOS as a file server...[16:44:09] <leecallen35> http://wiki.smartos.org/pages/viewpage.action?pageId=2687032[16:44:32] <leecallen35> I would appreciate getting edits from others, on both sides of the issue[16:44:39] <leecallen35> nahamu: this means you[16:44:54] * MerlinDMC leaving the office[16:44:57] <leecallen35> then in the future we can just paste this URL when anyone asks those questions[16:45:11] <nahamu> leecallen35: noted. :)[16:47:10] <phyrne> leecallen35: Great, reading now[16:49:03] <opeth__> I still would like to catch up with the zoned nfs-kernel-server[16:49:17] <opeth__> I loved it to bits when S11 came and changed my S10 world with that[16:49:37] <opeth__> this and zfs crypto are what still make me sigh every now and then[16:51:18] <phyrne> leecallen35: Very handy, nice one[16:51:28] *** nefilim has joined #smartos[16:51:36] *** potatosalad has quit IRC[16:51:49] <leecallen35> thanks. it needs help, starting with the stuff in red...[16:52:28] <leecallen35> but I am biased. does this accurately reflect the view for NOT using SmartOS as a NAS?[16:52:38] <phyrne> I'd love to contribute, but, considering I'm the one who raised the very same question earlier, prompting the creation of the document, I can't imagine I'd be all that much help ;)[16:52:47] <phyrne> leecallen35: Yes, I would say so[16:53:00] <leecallen35> jperkin: do you agree?[16:53:15] <phyrne> As you state that once virtualization is desired, SmartOS becomes a much more attractive candidate[16:53:20] *** khushildep has quit IRC[16:53:35] *** khushildep has joined #smartos[16:53:44] <jperkin> I'm probably not a good reviewer, I don't like the wiki ;)[16:53:52] <leecallen35> haha[16:54:08] <leecallen35> I dont think this page belongs where it is. it's just a temporary parking space.[16:55:06] <nahamu> On the server at $work, we now have a SmartOS box serving NFS and in-kernel CIFS.[16:55:19] *** jim80net has quit IRC[16:56:08] <leecallen35> so how is CIFS as a solution for windows clients? I played with it a long time ago, and I dont remember why exaclty I abandoned it, except the domain join persistence issue.[16:56:14] <nahamu> at boot time we run salt which updates the ldap and kerberos configs so that it looks like a pretty normal system.[16:56:23] <nahamu> we're using it in Workgroup mode[16:56:33] *** jim80net has joined #smartos[16:56:38] <phyrne> nahamu: Sweet. I'm getting jealous of all the people mentioning the use of SmartOS at work :( We only use Linux. The only Solaris we have is a Nexenta box :p[16:56:48] *** mary5030 has joined #smartos[16:56:49] <nahamu> we pull the NT hashes out of ldap and jam them into the smbpasswd file.[16:57:20] <nahamu> I haven't done a performance comparison with Samba (from what ira says I'd expect Samba to win), but it's relatively simple to get running[16:57:39] <leecallen35> nahamu: would you put in a paragraph or two describing that solution?[16:57:46] <leecallen35> yes I had the same conversation with ira[16:58:07] <leecallen35> but I don't know how much overhead lofs and samba add[16:59:00] <nahamu> might also be interesting to figure out if hyperlofs would have less overhead than lofs and write some goop to get the stuff mounted in via hyperlofs.[16:59:18] <nahamu> one thing that CIFS will get that samba will miss is the ability to expose the snapshots as Windows previous versions[16:59:25] <nahamu> (unless you run Samba in the GZ)[16:59:35] *** lloydde has joined #smartos[16:59:38] *** johnnuttall has joined #smartos[17:00:12] <nahamu> I'll try to carve out some time to add a description of my setup to that wiki page, but I might need a reminder next week to do it...[17:00:27] <leecallen35> you know, now I think I included too much detail on "how to". it's not intended as a how-to, just to describe an approach that works around the problems[17:01:07] <leecallen35> If I were to mentioned CIFS as an alternative, without all of your heroics, what would be the disadvantages?[17:01:28] <leecallen35> is there any access control?[17:01:36] <nahamu> depends on how many users.[17:02:24] <nahamu> on my home machine at boot time I jam in entries into /etc/passwd that correspond to entries in smbpasswd[17:02:48] <nahamu> (smbpasswd lives somewhere persistent)[17:03:15] <leecallen35> okay. give me a few minutes. I am going to reduce the level of "howto" detail and try to discuss both alternatives...[17:03:52] <leecallen35> maybe someday we will have how-to web pages for both approaches.[17:04:02] <nahamu> certainly an option.[17:04:22] <nahamu> high level overview page, and some subpages with howtos for given subsets of functionality.[17:04:44] *** W0rmDr1nk has quit IRC[17:04:59] *** lloydde has quit IRC[17:06:34] <nahamu> anyway, I have too much going on this week, but, leecallen35, ping me next week to add stuff to whatever page you want me to add to.[17:07:00] <leecallen35> sure. are you saying that CIFS references smbpasswd and /etc/passwd?[17:07:31] *** mamash has left #smartos[17:08:17] *** potatosalad has joined #smartos[17:09:31] <nahamu> /var/smb/smbpasswd is where the NT hashes get stored[17:09:41] <nahamu> those have associated usernames[17:09:58] <leecallen35> I get it[17:10:00] <nahamu> for those to mean anything in the filesystem, nss needs corresponding info[17:10:15] <nahamu> so either /etc/passwd (as I do at home), or LDAP (as I do at work) or whatever.[17:13:41] *** johnnuttall has quit IRC[17:13:42] *** wolfeida_ has joined #smartos[17:16:38] *** wolfeidau has quit IRC[17:16:55] *** johnnuttall has joined #smartos[17:23:49] <Jared_M> MerlinDMC: I don't see lz4 or xz on my SmartOS global zone, is it in a non-standard place?[17:24:25] <rmustacc> Jared_M: We don't ship it as part of the platform.[17:24:44] <Jared_M> no problem[17:25:01] <Jared_M> so gzip still the recommended way to make a img for imgadm?[17:25:16] <rmustacc> Assuming you're on new enough bits, the lz4 compression exists in zfs.[17:25:23] <rmustacc> Jared_M: I would use basic gzip, yeah.[17:25:30] <rmustacc> Pretty good tradeoff in terms of time/space.[17:25:40] <rmustacc> In our experience we've found bzip2 to not be worth it.[17:26:10] *** alcir has quit IRC[17:27:41] <Jared_M> no way I was going to try and bzip2 70gb's[17:27:52] <Jared_M> i was just wondering if there was some other obvious choice I was missing[17:27:58] <Jared_M> thanks rmustacc[17:28:45] <jperkin> we should probably ship pigz[17:29:02] <jperkin> (multi-core drop-in gzip replacement)[17:29:39] <leecallen35> If one installs software packages in the GZ, and later updates SmartOS (eg the USB key), might the packages someday be broken due to library mismatches?[17:31:41] <rmustacc> leecallen35: They should not.[17:31:59] <rmustacc> Though of course there are probably some edge cases that could happen.[17:32:11] <rmustacc> But in general we try to make sure that the platform is always backwards compatible with respect to existing software.[17:32:18] <rmustacc> Even when that really sucks.[17:32:20] <leecallen35> wow. that seems really difficult.[17:32:34] <rmustacc> Well, it means you have to be careful about what's a committed and stable ABI.[17:32:57] <rmustacc> There are some things in the platform that have compilation symlinks that aren't stable. If you're using it (say libzfs) then we're alloewd to break that.[17:33:04] <JT-EC> pigz is quite good but does generate huge loads on the host machine (may just be our flags/options) so may need to be careful if you create images from production machines on those production hosts.[17:33:11] <leecallen35> rmustacc: thanks[17:33:19] <rmustacc> But something like libc, libm, etc. you shouldn't.[17:33:41] <rmustacc> This is also why we're increasingly basically making it impossible to use the platform dependents, eg. the platform's private copy of openssl.[17:33:48] <rmustacc> Because we want to make sure that we can't break a user on upgrade.[17:33:55] <rmustacc> It also means that we have to live with our mistakes for some time.[17:34:25] <leecallen35> I understand[17:34:29] <leecallen35> thanks[17:34:52] <rmustacc> We may hit a mystical point where we eventually say the backwards compat copies of certain private libraries (openssl, libxml2, etc.) may go away and if you got on a very very old platform where compilation symlinks existed then that'll stop working.[17:34:59] *** chrispercol has joined #smartos[17:35:51] <jperkin> and at the same time we're working on eradicating any linkage against those libraries from pkgsrc[17:36:21] <chrispercol> hi guys, I'm having issues doing a git clone from the smartos /opt directory...I have no issue using my private key to access the host, are there some little SmartOS gotchas I'm not aware of doing this?[17:36:25] <rmustacc> I'd like to hope that you're more likely to break upgrading across pkgsrc repositories because third party software likes to break itself.[17:36:50] <rmustacc> chrispercol: Are you doing this in a zone, GZ? I don't know of any issues.[17:37:09] <leecallen35> rmustacc: so this is not a case against installing software packages in the GZ. I had always assumed it was.[17:37:39] <rmustacc> leecallen35: That specific problem also exists in zones because we use sparse zones.[17:37:55] <chrispercol> I setup authorized keys and am have setup /opt in the GZ as a git repo which I am trying to clone to another machine[17:37:57] <leecallen35> right[17:38:50] <chrispercol> I'm pretty famililar with ssh and git, but not on SmartOS[17:38:53] *** johnnuttall has quit IRC[17:38:59] <rmustacc> chrispercol: They should be no different.[17:39:05] <chrispercol> it's almost like I am having permissions problem[17:40:04] <chrispercol> rmustacc: cheers, I'll keep plugging away[17:40:24] <rmustacc> Aside from the caveat that in the GZ root's home directory is not persistent.[17:42:04] <chrispercol> yeah, I rebooted after setting up /usbkey/config, all was good with auth keys[17:49:18] *** mikl has left #smartos[17:53:48] <phyrne> Whelp, I'm off for now. Thanks for your help guys, I'm sure I'll be back soon :)[17:53:52] <phyrne> o/[17:53:52] *** phyrne has quit IRC[18:05:02] *** CaTtleyA has quit IRC[18:05:08] *** chorrell has joined #smartos[18:05:55] *** CaTtleyA has joined #smartos[18:06:36] *** CaTtleyA has quit IRC[18:07:12] *** CaTtleyA has joined #smartos[18:13:39] *** johnnuttall has joined #smartos[18:15:15] *** lloydde has joined #smartos[18:15:18] *** dap1 has joined #smartos[18:15:56] <chrispercol> back again...I know why I am having git problems but not sure howto fix...basically /opt/local/bin is not in the path for my remote clients, can anyone think of a way around?[18:17:31] <chrispercol> that is, if I do export $PATH in smartos I get the path to /opt/local/bin but when I do it via ssh I only see /usr/sbin and /usr/bin[18:17:58] <jperkin> you need a login shell to get the PATH setup from .profile[18:18:16] <jperkin> so 'ssh foo; git blah' will work, but 'ssh foo "git blah"' will not.[18:19:42] <chrispercol> jperkin: thanks, is .profile something I do from /usbkeys/config?[18:20:07] <jperkin> no. are you trying to automate logging in and running git all in one go?[18:20:17] <jperkin> if so 'ssh foo ". .profile; git blah"'[18:20:39] <chrispercol> yes, git clone ssh://user@soemthing[18:21:22] <chrispercol> should say my remote is windows[18:21:30] <Alasdairrr> two kernel panics on SmartOS boxes running our latest images recently[18:21:53] <wesolows> illumos bug ID?[18:22:13] <Alasdairrr> Still gathering information and looking at things[18:22:42] <Alasdairrr> I'll grab the stack trace, sadly the second one didn't complete the dump as the dump device was too small[18:23:48] <Alasdairrr> http://paste.ec/?b4efe4b1534d960d#5rIgv6JM2HF29Roq7EAbj5WmbrKUjmTaXa/lNJDzMgw=[18:23:57] *** ryancnelson has joined #smartos[18:24:09] <Alasdairrr> This was the second one on a different box: http://linux01.everycity.co.uk/~aszeszo/skitched-20130925-175322.png[18:26:41] <wesolows> I've not seen either of those before. You seem to be in a path where we're generating ereports for a bad block or bad disk though.[18:26:47] <wesolows> That might not be well-tested.[18:27:05] *** CaTtleyA has quit IRC[18:27:05] <wesolows> Seems likely to be a dumb bug where we're feeding and nvlist routine NULL.[18:28:53] <Alasdairrr> Yeah[18:30:03] <Alasdairrr> Both machines were scrubbing at the time[18:31:53] <Alasdairrr> It's a bit of a confused issue[18:31:56] <wesolows> I've been looking at this area recently but haven't changed anything in the zio_done path or zfs_fm.c. We will likely need to for other reasons but I don't see any recent change that's a likely cause.[18:33:15] <Alasdairrr> On box so1-7 we suffered a harddrive failure (Dell Savvio 10k.5 600GB SAS with an older firmware) which locked IO up such we had to reboot. The LSI MPTBIOS got stuck at detecting drives, so we had to physically yank the drive to get back. Because we netboot the box booted back up on the latest image. There were checksum errors so (I think) a zpool clear triggered a scrub and then a bit later the box panic'd[18:34:23] <Alasdairrr> The box that panic'd today I need to speak to aszeszo about, I don't know how it came to be doing a scrub, but it was on the latest image too[18:34:42] <Alasdairrr> Seems very strange for two boxes to hit the same bug in such close proximity[18:36:09] *** d[^_^]b has joined #smartos[18:39:50] <Alasdairrr> We don't run the vanilla smartos, I think we have Nexenta's mpt_sas patches applied. Fat lot of help they were with that drive failure. Perhaps we need to back those changes out and wait for the good stuff to arrive...[18:40:11] *** johnnuttall has quit IRC[18:45:11] *** mamash has joined #smartos[18:46:20] *** chrispercol has quit IRC[18:52:29] <Alasdairrr> http://paste.ec/?9f84f6f8a6a862df#mUHAeaAdq8wsWwusQ7RrXZ+MI+l1r0c/VJT0mApNLeY=[18:52:36] *** nefilim has quit IRC[18:53:07] <Alasdairrr> incredibly slow resilver time too[18:53:14] <Alasdairrr> Evacuation time for both these boxes I think[18:53:20] *** KermitTheFragger has quit IRC[18:58:27] <wesolows> Why would you ever enable loading the oprom on an LSI controller?![18:59:16] *** dap1 has quit IRC[18:59:25] <wesolows> oproms are needed/useful only for booting, and booting from disk is not supported.[18:59:25] <Alasdairrr> It comes enabled by default. Don't worry, we have made a note to our provisioning checklist to ensure it gets disabled in future[18:59:35] <wesolows> heh, ok. good move for sure.[18:59:51] <Alasdairrr> :-)[19:02:39] <Alasdairrr> wesolows: What's your opinion of the nexenta timeout patches? Should we just pull them all out and revert to SmartOS generic?[19:02:49] <Alasdairrr> I saw you recently changed some sd timeouts[19:03:07] <wesolows> I've struggled to review their changes because the webrevs all seem to be cumulative with the LED control removal as a base.[19:03:46] <wesolows> FWIW, though, I believe that our current platform allows you to set every timeout that needs changing using a combination of /etc/system and driver.conf files.[19:03:55] <wesolows> And in fact our platform has them all set sanely.[19:04:25] <wesolows> So if there are still timeout-related problems that are not addressed by our current platform as shipped, I'd like to get bugs filed for them.[19:04:26] <Alasdairrr> "set sanely" = https://github.com/joyent/illumos-joyent/commit/1c485ea3760aeec397cb5b579d3feaf1c097ee4b ?[19:04:40] <wesolows> That piece just enabled setting them.[19:05:08] <wesolows> The more relevant one was https://github.com/joyent/smartos-live/commit/e07b2d7fd21c58c9a366d7a5ae040a212f850af5[19:05:37] <Alasdairrr> Sure[19:06:23] <Alasdairrr> So in theory with those in place, the nexenta mpt_sas patches in theory shouldn't be needed, unless there are unknown bugs that need filed[19:09:07] <wesolows> As far as I know.[19:09:19] <wesolows> I haven't been able to pick apart their changes.[19:09:40] *** tonist has joined #smartos[19:09:48] <Alasdairrr> There was one specific change that Albert Lee wrote that introduced a kill_target routine[19:10:12] <wesolows> Which one of http://kebe.com/~danmcd/webrevs/mptsas-webrevs/ is it?[19:10:18] * Alasdairrr looks[19:10:28] <Alasdairrr> http://kebe.com/~danmcd/webrevs/mptsas-webrevs/mptsas-4/mptsas-4.patch[19:10:31] <wesolows> it looks like those have been split up nicely now[19:10:48] <wesolows> oh, that one.[19:10:53] <Alasdairrr> if a timeout was hit 2 times within 30 seconds it'd forcibly nuke the SCSI target[19:10:57] <wesolows> I hate that change and will fight it tooth and nail.[19:11:17] <wesolows> If you want to do that, you generate ereports for timeout events and have fmd retire the device.[19:11:25] <wesolows> You don't shove FMA into a drive.[19:11:27] <wesolows> *driver[19:12:05] <wesolows> I also happen to think they're being way too aggressive, but that's a policy choice -- one that in fact is much easier to make in userland, in an FMA DE's .conf file.[19:12:20] <Alasdairrr> Sure, it's a hack. But without it, what is there already to achieve the same result? Nothing? Or should Nexenta have just tuned the sd stuff you sorted out in the two aforementioned changesets?[19:12:40] <wesolows> There are two different things going on here.[19:12:59] <wesolows> First, what constitutes a timeout, and how many times should the SCSI framework retry when they happen?[19:13:24] <wesolows> Second, when timeouts or other errors occur, what happens to identify and isolate ("diagnose" and "retire") the failing device?[19:13:30] <wesolows> My change affects only the first.[19:13:47] <wesolows> Theirs is trying to improve the second, but in a horribly over-aggressive way and in the worst possible place.[19:14:02] <wesolows> If that change goes into illumos, we will back it out.[19:14:11] <Alasdairrr> Sure[19:14:18] <Alasdairrr> The thing is, we were repeatedly hit by nearline SAS drives "failing" but not being timed out, causing all IO to hang, and were a bit at our wits end.[19:14:34] <wesolows> We already generate ereports in sd. How fucking hard is it to write a DE that acts on them?[19:14:51] <wesolows> We occasionally see byzantine disk failures too.[19:14:55] <Alasdairrr> We've now had our first drive failure leading to a lockup on SmartOS on our Dell C6100 kit and thats *with* that nexenta code in[19:15:15] <Alasdairrr> It's like, when is this ever going to be addressed properly?[19:15:20] <wesolows> We'd like to fail things that time out repeatedly too (although even with very aggressive timeouts like 5s, it's often the case that these devices actually *don't* time out at all).[19:15:22] <Alasdairrr> We've already switched to NetApps for NFS storage[19:15:33] <wesolows> When people stop fucking hacking mpt_sas and start doing the right fucking thing.[19:15:48] <wesolows> And the best way to encourage that is to categorically reject this patch.[19:16:21] <wesolows> FMA is the answer here. If we collectively don't believe we can ever make FMA useful and reasonably complete, then we need to come up with a viable alternative to it.[19:16:46] <wesolows> I don't think a thousand points of per-driver mdb-hackables is better than FMA.[19:17:01] <wesolows> So if it's not FMA, it needs to be something else, but definitely not this.[19:17:15] <Alasdairrr> I doubt anyone will disagree with you, I'm not fighting to get that patch into illumos :-)[19:17:56] <wesolows> Having FMA in place but not using it is actually the worst of all worlds -- we struggle mightily to get our Operations team to use FMA, but when it doesn't diagnose a large fraction of failures, they just go write their own scripts instead with policies we can't see or fix.[19:18:23] <wesolows> And of course a completely opaque driver-specific "diagnosis" here is worse still since the device just "goes away".[19:18:46] <wesolows> Anyway... sigh.[19:19:14] <Alasdairrr> Okay. Last question, is it a lot of work to address this and is it something you reckon Joyent would ever look at or address, or something that's not an issue right now[19:19:30] <Alasdairrr> I feel like we're the only company suffering from hosts locking up when drives fail[19:19:35] <wesolows> It's a medium-sized project. A few months for one person, with no distractions I think.[19:19:43] <Alasdairrr> Whoa okay, that's a lot of work[19:20:08] <wesolows> And yes, I want to address it. It's one of the options I'm going to offer for something to focus on over the next 1-3 quarters.[19:20:32] <wesolows> Whether we will do it, I can't say. I don't make any decisions here.[19:20:49] <Alasdairrr> Sure, well what you've said so far is really appreciated[19:23:15] *** wolstena has joined #smartos[19:29:43] *** khushildep has quit IRC[19:43:31] *** texarcana has quit IRC[19:45:54] *** texarcana has joined #smartos[19:48:30] *** mamash has left #smartos[20:01:27] *** jaimef has quit IRC[20:07:12] *** jaimef has joined #smartos[20:08:43] <Jared_M> question about memory, I just got this message: "VM "ram" value: 8192 is greater than available memory: 2987" when doing a vmadm create.[20:08:44] <Jared_M> [root@00-25-90-96-33-fa /zones/config_files/clones]# prtconf | grep -i memory[20:08:45] <Jared_M> Memory size: 32739 Megabytes[20:09:01] <Jared_M> and my current VM's only add up to 16GB[20:09:02] *** lloydde has quit IRC[20:09:05] <Jared_M> so what gives?[20:09:35] *** lloydde has joined #smartos[20:09:41] <leecallen35> nahamu: I am pretty much done with that wiki page http://wiki.smartos.org/pages/viewpage.action?pageId=2687032[20:09:58] <leecallen35> please review it when you can. I reduced the level of detail to the point there can't be too much wrong.[20:11:34] <leecallen35> hmmm what did I do wrong that it has such a funky URL?[20:11:49] *** ryancnelson has left #smartos[20:13:42] *** tonist has quit IRC[20:13:45] *** lloydde has quit IRC[20:13:56] *** d[^_^]b has quit IRC[20:14:34] *** lloydde has joined #smartos[20:16:22] *** d[^_^]b has joined #smartos[20:19:19] *** andy_js has quit IRC[20:26:44] <sebasp_> Jared_M: See this http://wiki.smartos.org/display/DOC/SmartOS+Command+Line+Tips#SmartOSCommandLineTips-Checkingtheamountof%26nbsp%3Bmemoryonasystem[20:29:44] *** andoriyu has joined #smartos[20:29:50] <Jared_M> so, looking at ::memstat, what is "kernel" versus "anon"?[20:30:38] *** sebasp_ is now known as sebasp[20:31:28] <Jared_M> seems weird to me the kernel is using 12gb of ram[20:31:46] <Jared_M> and 'anon' is using another 12[20:31:58] *** FatDarrel has joined #smartos[20:32:19] <MerlinDMC> Jared_M, how big is your zfs ARC?[20:34:10] <MerlinDMC> Jared_M, also maybe you're somehow still hitting this: https://www.illumos.org/issues/1974[20:34:20] <Jared_M> MerlinDMC: [root@00-25-90-96-33-f4 /zones/config_files/clones]# kstat -p zfs:0:arcstats[20:34:20] <Jared_M> zfs:0:arcstats:c 16671102797[20:34:20] <Jared_M> zfs:0:arcstats:c_max 33246859264[20:36:12] <sebasp> Jared_M: seems your not caping your arc[20:36:47] <Jared_M> so "kernel" is taking up 12gb, "zfs file data" is 8gb, "anon" is 10gb, seems high when I have only 10gb worth of kvm and 6gb worth of smartos guests[20:37:08] <Jared_M> by default does smartos just set the max to the maxium available ram?[20:38:00] <wesolows> ZFS in general will use as much memory as ARC as it can, minus a small amount[20:38:14] <Jared_M> good to know[20:38:17] <wesolows> there are also a couple of mechanisms that will cause it to shrink when there's memory pressure.[20:38:51] <wesolows> However, an attempt to lock down a giant chunk is not "memory pressure", so you often end up here. There are workarounds to induce shrinking before creating large KVM instances.[20:39:04] <rmustacc> Most likely what you're seeing is this which hasn't made it into a release yet: https://github.com/joyent/smartos-live/commit/895a4ff0a9bc6a31ff865c8505e4cc1382d5d876[20:39:26] <wesolows> Or that.[20:39:42] <wesolows> It does help to debug these things to root cause. :)[20:42:26] <Jared_M> so is the only option i have right now to seriously downsize the arc, run vmadm create, and then increase the arc back up to 8gb?[20:42:47] <nahamu> the arc will grow on its own[20:43:10] <nahamu> you just need to provide memory pressure to shrink it enough that QEMU can lock the RAM it needs.[20:43:12] <Jared_M> that accounts for the 8gb in "zfs file data", but what else is in the "kernel" category that could account for 6gb[20:43:26] <opeth__> jperkin: I like the way pkgsrc's patch target applies everything under 'patches' called patch- given the mps target has also been made already... I've just applied the 3rd party vda patch seamlessly.[20:43:48] <opeth__> to postfix, that is[20:43:55] <Jared_M> nahamu: but how do I size it down w/o changing its max? if I adjust its max I would hope it didn't grow past that[20:45:01] <szaydel> wesolows: Cannot agree with you more about mpt_sas, and timeout settings. The changes you linked to seem way more sane, though it seems that even 10s is way long. I would guess that one should give up after 5 seconds or so, which in itself seems like an eternity.[20:45:49] <sebasp> Jared_M: why would you not max it?[20:45:57] <wesolows> szaydel: It probably is, but there's a catch: the timeout applies as soon as the command is sent. If you're at the back of a 10-deep (or 32-deep, for non-ZFS) queue, it's not hard for the last guy to time out if the ones in front of him are slow.[20:45:58] <sebasp> Jared_M: also look at this http://www.solarisinternals.com/wiki/index.php/ZFS_Evil_Tuning_Guide#Determining_ARC_memory_consumption_and_other_related_stats[20:46:25] <wesolows> I think we did experiment with 5s at Fishworks, though. It's not a terrible idea.[20:46:28] <szaydel> wesolows: That makes sense I suppose.[20:46:28] <nahamu> manual workaround: https://www.illumos.org/issues/1974[20:46:50] <szaydel> Seems like your queue should be less if items wait that long! :)[20:46:57] <wesolows> Surely it would help to figure out whether this is actually caused by ARC growth rather than the bug rmustacc pointed out, before doing other stuff?[20:47:18] <nahamu> wesolows: by all means. I defer to you. :)[20:47:42] <wesolows> szaydel: yes, probably. Although it's hard to know why they're long; in general they shouldn't be. We usually expect to get EIO after about 3s anyway.[20:48:03] <szaydel> wesolows: Yep, that's what I'd expect.[20:48:15] *** andy_js has joined #smartos[20:48:26] <nahamu> oh, I missed the initial error message. I didn't realize it came from vmadm.[20:49:35] *** lloydde has quit IRC[20:49:38] <Jared_M> wesolows: I will run any command that would be of interest to you[20:50:11] *** lloydde has joined #smartos[20:50:12] <wesolows> I'd think the approach here would be to copy that file into /var/tmp, apply the delta, then lofs-mount it over and see if that fixes it for you.[20:50:31] <wesolows> If it does, you've hit OS-2497 and can safely drive on.[20:51:03] <Jared_M> which file?[20:51:29] <nahamu> would pulling down the latest version of the file work?[20:51:32] <nahamu> https://raw.github.com/joyent/smartos-live/895a4ff0a9bc6a31ff865c8505e4cc1382d5d876/src/vm/node_modules/VM.js[20:51:46] <wesolows> /usr/vm/node_modules/VM.js I think[20:51:50] <wesolows> yes[20:52:45] <nahamu> cd /tmp; curl -kO https://raw.github.com/joyent/smartos-live/895a4ff0a9bc6a31ff865c8505e4cc1382d5d876/src/vm/node_modules/VM.js ; <something like "moun -F lofs /tmp/VM.js /usr/vm/node_modules/VM.js>[20:52:50] <nahamu> *mount[20:53:38] <Jared_M> will try[20:54:02] <nahamu> I'm not certain of the mount command, but it's something vaguely like that.[20:54:31] <wesolows> it's that.[20:54:36] *** lloydde has quit IRC[20:54:46] <nahamu> \o/[20:55:28] <nahamu> I guess the illumos stuff is sinking nicely into my brain. :)[20:56:30] <Jared_M> okay, that worked, it didn't complain about ram anymore, it just failed on something else, but I can search for that[20:59:09] *** bens1 has quit IRC[20:59:30] *** sebasp is now known as sebasp_[21:01:09] *** gkyildirim has joined #smartos[21:10:13] *** gkyildirim has quit IRC[21:19:42] <emsii> Hi. I'm planning on building an Atom S1260 based home server with SmartOS.[21:19:42] <emsii> S1260 has VT-x, but does not have EPT.[21:19:43] <emsii> Have I understood correctly that KVM on SmartOS will only work on processors with EPT?[21:21:36] *** lloydde has joined #smartos[21:22:20] *** nikolam has joined #smartos[21:26:14] <jesse_> yes and no[21:26:46] <jesse_> I think the unofficial amd-supporting kvm also supports non-ept intel cpus[21:27:57] <konobi> a fairly recent core i3 will work well[21:28:45] *** lloydde has quit IRC[21:30:32] <nahamu> konobi likes his new NUC. :)[21:31:13] <richlowe> shame they only take one disk[21:43:58] *** ryancnelson has joined #smartos[21:50:08] *** julianduque has joined #smartos[21:59:02] *** sebasp_ is now known as sebasp[22:04:07] *** lloydde has joined #smartos[22:10:32] *** gemelen has joined #smartos[22:13:52] *** leecallen has joined #smartos[22:13:59] <Jared_M> so I tried to figure this out on my own once my ram issue was solved, but I'm having an off day, I can't get a image from imgadm to be used as the disk to a new vm[22:14:18] <Jared_M> let me gist the output[22:14:22] <Jared_M> pasting it was too much[22:14:45] <MerlinDMC> Jared_M, and gist your json as well[22:15:21] <Jared_M> will do[22:16:22] <Jared_M> https://gist.github.com/jaredmorrow/e8389237d6dbec4c8c82[22:16:23] *** leecallen35 has quit IRC[22:16:48] <Jared_M> I followed the guide here: http://wiki.smartos.org/display/DOC/Managing+Images[22:17:14] <MerlinDMC> ah so your own image?[22:17:27] <Jared_M> yes[22:18:15] <Jared_M> imgadm imported it, so I thought I was successful in creating it[22:18:18] *** mamash has joined #smartos[22:18:59] <MerlinDMC> Jared_M, replace the "size" property with "image_size" in your json[22:20:44] <MerlinDMC> and I don't know if you had image_size in your dsmanifest ... that should have been in there[22:20:53] <Jared_M> I did[22:21:06] <Jared_M> that was the size of the .gz on disk[22:21:41] <MerlinDMC> image_size should have been the size of the disk you configured while creating the vm you used to create the image from[22:21:45] <Jared_M> actually, no it was size[22:21:49] <Jared_M> "files": [[22:21:50] <Jared_M> {[22:21:50] <Jared_M> "path": "riak-test-centos-6.zvol.gz",[22:21:50] <Jared_M> "sha1": "55b1df32d2cc5da61ebda51b05c951070f0651ed",[22:21:50] <Jared_M> "size": "3613142729"[22:21:56] <Jared_M> that was in the dsmanifest[22:22:21] <Jared_M> following exactly http://wiki.smartos.org/display/DOC/Managing+Images[22:23:02] <MerlinDMC> Jared_M, https://images.joyent.com/images/1327ed36-5130-11e2-95a8-9b5a153adf3e ... I'm talking about the 4 items on the bottom ... nic_driver, disk_driver, cpu_type and image_size[22:23:05] <ryancnelson> he's saying, change it in your json.[22:23:37] <MerlinDMC> ryancnelson, I also think his dsmanifest might not been complete as well[22:23:40] <ryancnelson> sorry, "json" is ambiguous. in your "this-specific-provision-of-this-vm" json[22:23:49] *** sebasp has quit IRC[22:23:57] *** sebasp has joined #smartos[22:23:58] <ryancnelson> everything is json around here :)[22:24:22] <Jared_M> yeah, I changed it and it still didn't work[22:24:39] <Jared_M> "disks": [[22:24:39] <Jared_M> {[22:24:39] <Jared_M> "image_uuid": "29d60ba8-fb31-4055-be06-6428a727f68a",[22:24:39] <Jared_M> "boot": true,[22:24:40] <Jared_M> "model": "virtio",[22:24:40] <Jared_M> "image_size": 40960[22:24:40] <Jared_M> }[22:24:42] <MerlinDMC> riak-test-centos-6-2.json Line 14 should be image_size[22:25:10] <MerlinDMC> can you gist the dsmanifest you used to import your image?[22:26:15] <Jared_M> added https://gist.github.com/jaredmorrow/e8389237d6dbec4c8c82[22:26:51] <MerlinDMC> Jared_M, that one is missing at least 4 entries: nic_driver, disk_driver, cpu_type and image_size[22:27:03] <Jared_M> well then the wiki page is out of date[22:27:13] <MerlinDMC> where the image_size might be your biggest problem because that is verified afaik[22:27:51] <Jared_M> I will happily follow updated documentation if you have a link?[22:28:34] <Jared_M> also imgadm install doesn't verify it[22:28:47] <Jared_M> b/c imgadm installed via that manifest fine[22:29:05] *** sebasp has quit IRC[22:29:13] *** sebasp has joined #smartos[22:29:35] <MerlinDMC> Jared_M, imgadm destroy and install again with something like this: https://gist.github.com/MerlinDMC/9807993f5903e931b0dd[22:29:43] *** bens1 has joined #smartos[22:29:48] <MerlinDMC> the wiki page seems to have no example for a kvm manifest[22:32:49] <ryancnelson> as i understand it, either the manifest, or your json need to have that info.[22:33:02] <ryancnelson> it's valid without, but if neither have it, you get an error[22:33:46] <MerlinDMC> mibe hates me ... guess I have to get some sleep[22:34:06] *** bens1 has quit IRC[22:34:14] <ryancnelson> we should probably update the wiki with an example uuid that's still published in https://images.joyent.com/images[22:34:23] *** chorrell has quit IRC[22:35:22] <MerlinDMC> also with imgadm v2 the manifest is a lot lighter ... maybe we should change that to the new format as well then[22:37:10] <Jared_M> a working end-to-end example would have saved me hours[22:37:58] <rmustacc> Jared_M: Sorry we don't have one. Perhaps can you create one from what you have and add that in its place?[22:39:21] <Jared_M> does the wiki take outside submissions now?[22:39:40] <MerlinDMC> the smartos wiki can be edited if you create an account[22:40:50] <rmustacc> Yeah, certainly.[22:40:51] *** harmw has quit IRC[22:41:16] <Jared_M> okay, for some reason I thought I tried at one point, but couldn't edit[22:41:26] <Jared_M> I'm importing the image to imgadm again[22:41:44] <Jared_M> if it works, I'll write all this down, VM.js patch included[22:42:03] <rmustacc> What was the patch to VM.js?[22:42:24] <MerlinDMC> OS-2497[22:42:55] <rmustacc> Oh, no need to write that down as that'll just be part of the next release.[22:46:18] *** harmw has joined #smartos[22:46:34] *** mamash has left #smartos[22:51:38] <Jared_M> well that destroy and reinstall didn't work[22:51:39] <Jared_M> [root@00-25-90-96-33-fa /zones/isos]# imgadm install -m riak-test-centos-6.dsmanifest -f riak-test-centos-6.zvol.gz[22:51:40] <Jared_M> installing image 29d60ba8-fb31-4055-be06-6428a727f68a[22:51:40] <Jared_M> Error: {"error":{"code":21,"message":"error renaming image"}}[22:53:11] <Jared_M> smh[22:53:18] <MerlinDMC> mostly means that destroy wasn't successful - happened here sometimes when I had a provisioned machine not deleted before destroying the image ... zfs destroy was my friend[23:01:22] *** jaimef has quit IRC[23:01:27] *** wolfeida_ is now known as wolfeidau[23:01:42] *** jaimef has joined #smartos[23:12:10] <Jared_M> MerlinDMC: so imported image again with your manifest[23:12:28] <Jared_M> create failed, but I hadn't added anything there, should the image_size be in the disks section, or outside of it?[23:12:48] <Jared_M> right now i just have s/size/image_size/[23:13:10] <MerlinDMC> in the dsmanifest it should be in the root, in the json that you feed into vmadm create it should be in the disk section[23:14:01] <Jared_M> okay, it is still giving me the same error[23:15:37] <ryancnelson> compare it to something like this?[23:15:37] <ryancnelson> https://images.joyent.com/images/1327ed36-5130-11e2-95a8-9b5a153adf3e[23:16:41] <Jared_M> well my dsmanifest imported with https://gist.github.com/MerlinDMC/9807993f5903e931b0dd fine[23:17:22] <Jared_M> my new vm's json is this https://gist.github.com/jaredmorrow/e8389237d6dbec4c8c82#file-riak-test-centos-6-2-json[23:19:56] <MerlinDMC> Jared_M, I hae no kvm images here - I will be in the office in about 7h there I could check if provisioning works for my images with the latest smartos release[23:20:02] * MerlinDMC need some sleep[23:20:16] <Jared_M> okay, thanks. I need a break too[23:20:21] <Jared_M> thanks for the help[23:22:23] *** nikolam has quit IRC[23:30:35] *** andoriyu has quit IRC[23:30:53] *** andoriyu has joined #smartos[23:33:03] *** jim80net has quit IRC[23:53:07] *** andy_js has quit IRC[23:57:39] *** wolstena has quit IRC