[00:26:25] * haba has AFS server which had sudden SCSI problem with its RAID box. No fun. Now after reboot everything seems fine. Grrr. [00:32:25] --- Russ has left: Disconnected [00:38:53] --- dev-zero@jabber.org has left [01:50:10] --- dev-zero@jabber.org has become available [02:31:49] --- Simon Wilkinson has become available [02:36:37] --- Simon Wilkinson has left [03:31:17] --- haba has left [03:36:36] --- Simon Wilkinson has become available [04:11:30] --- meffie has become available [04:26:19] --- Simon Wilkinson has left [04:44:43] --- jaltman has left: Disconnected [05:07:00] --- haba has become available [05:17:01] Today I got in my Volserlog: Fri Nov 20 13:51:16 2009 1 Volser: DumpVolume: Rx call failed during dump, error -01 And I have trouble backing up / dumping certain volumes from that server. Ideas? Suggestions? [05:29:17] what version? what's on the other end? [05:59:59] --- Jeffrey Altman has become available [06:03:54] --- Simon Wilkinson has become available [06:04:36] --- Simon Wilkinson has left [06:12:38] 2.4.11 volser and 2.4.10 vos on amd64 [06:14:53] upgrade vos. [06:15:00] OK [06:15:41] yay, finally someone who doesn't ask for help and then bitch when you try to offer it! [06:15:53] (it'll be the rx idle time issue in 1.4.10 i bet) [06:17:21] I will not bet against, building RPM now :) [06:20:38] vos dump bla bla.... -verbose Incremental Dump (as of (null))... null?! [06:21:20] Btw is -time 11/18/2009 the same thing as -time 11/18/2009\ 00:00 ?? [06:21:34] * haba does not know when "12:00 am" is. [06:22:48] i think so [06:22:52] Then one should test that 11/18/2009\ 00:00 is the same thing as 11/17/2009\ 24:00 [06:46:59] --- abo has left [06:47:06] --- abo has become available [07:05:35] --- dev-zero@jabber.org has left [07:05:48] --- dev-zero@jabber.org has become available [07:14:34] --- deason has become available [07:19:04] since the change of vos, I did not have a problem dumping a volume yet. [07:19:44] dude, i didn't give you a reason and you said "sure, i'll upgrade one minor version" without a question. you made my week. [07:21:26] I have a RAID box which failed a HD, then recovered to a spare which now reports bad blocks and I'm moving out from at the same time as I am migrating my TSM to another server. My week is allready -ehm- "fine". [08:02:29] --- meffie has left [08:34:03] --- jaltman has become available [08:41:37] --- dev-zero@jabber.org has left [08:41:50] --- jaltman has left: Disconnected [08:41:53] --- dev-zero@jabber.org has become available [08:52:15] --- meffie has become available [08:56:14] idle question... would it be feasible/desirable to have a mailing list or something to get notifications of new gerrit changes? [09:32:12] i was just discussing at lunch how i'd like to see that. [09:33:18] i'm already happy that we encountered the bug fixed in 619 mere hours before it was merged into openafs-stable-1_4_x. there are a lot of git/gerrit fans here. [09:33:21] You can tell gerrit to send you notifications of new changes. [09:34:53] i wanted jabber to get it [09:35:07] not this muc. but *a* muc [09:35:58] jhutz++ [09:36:03] i did not notice that feature. [09:42:24] oh wow, I didn't notice that; thanks jhutz [09:49:07] --- reuteras has left [09:50:45] i'm i right in that comments need to be responded to in the gerrit web interface, ie, i should not just reply to the mail. [09:51:15] s/i'm/am/? [09:51:21] but you are correct [09:52:55] er, yeah, s/i'm/am/. thanks. [10:35:47] --- Russ has become available [10:38:49] --- dev-zero@jabber.org has left [10:54:23] --- jaltman has become available [11:06:12] --- jaltman has left: Replaced by new connection [11:06:18] --- jaltman has become available [11:17:10] --- jaltman has left: Disconnected [11:40:59] --- dev-zero@jabber.org has become available [11:46:12] --- jaltman has become available [11:50:33] --- haba has left [11:56:20] --- dev-zero@jabber.org has left [11:56:32] --- dev-zero@jabber.org has become available [12:07:46] --- jaltman has left: Disconnected [12:07:54] --- jaltman has become available [12:47:32] --- dev-zero@jabber.org has left [13:21:47] --- dev-zero@jabber.org has become available [13:22:01] --- Simon Wilkinson has become available [13:24:16] There's an open issue on gerrit to accept comments by email, but that's not be done yet. [13:27:59] --- haba has become available [13:51:04] --- dev-zero@jabber.org has left [13:51:19] --- dev-zero@jabber.org has become available [14:13:14] --- meffie has left [14:21:25] --- mdionne has become available [14:36:51] --- Jeffrey Altman has left [14:43:00] --- mdionne has left [14:45:26] --- mdionne has become available [14:45:39] Simon: commit 4057a994 introduces some warnings - I think the void * trick was there because UFSOpen and MemCacheOpen share a prototype because of afs_cacheOps [14:46:21] Oh. Yuck. [14:46:27] That sounds right now that you mention it. [14:47:10] So i'm not sure what the cleanest fix is at this point [14:47:39] I think probably just revert that change. [14:48:16] We should probably comment somewhere that that's why we're returning a void * in that case. [14:48:55] making MemCacheOpen return osi_file * would not be quite right [14:49:05] The only other option would be to make the open functions return a union, but that does make our cache structures less pluggable. [14:49:09] yes, a comment somewhere would be useful [14:49:50] Do you want to push a revert into gerrit, or shall I? [14:50:15] Doesn't matter - I can push it. [14:50:20] Cool. Thanks. [14:50:46] If you fancy pushing another change with a suitable comment - probably next to the cacheOps prototype, that would be good, too! [14:51:15] --- abo has left [14:51:51] --- abo has become available [14:53:00] --- stevenjenkins has left [14:54:06] Pushed [14:54:54] Thanks. [14:55:57] --- stevenjenkins has become available [14:58:16] Hopefully if I get some spare online moments over the next week, I can get the buildbot code finished. Then, hopefully we can catch more of these "works on X but warns on Y" problems before they hit the tree. [15:00:02] --- dev-zero@jabber.org has left [15:00:15] Yes that would be good. Some commits look obvious but when it builds we get surprises [15:01:05] Indeed. And I'm bad that when I'm working offline, I tend to just verify on Darwin, then, if I think the change is trivial, push it. [15:01:13] The problem is that some of the trivial changes aren't. [15:01:36] Would the tests get run at the intial submit, or when it's accepted and pushed to the master branch? [15:01:45] They'd get run at initial submit. [15:02:01] (or whenever build time becomes available in the build farm) [15:02:07] Ok, if it covers several platforms it would great [15:02:29] We'll cover as many platforms as we can get cycles donated for ... [15:02:58] Covering the main ones would already be a big step forward [15:03:04] We should be able to do RHEL, Fedora and OpenSolaris based on what I know is available at the moment. [15:03:45] .. and Darwin I would assume? [15:03:58] Actually, not currently :( [15:04:51] I need to find someone with Mac OS server hardware that I can lean on ... [15:05:06] (oh, and IRIX 6.5) [15:05:14] --- haba has left [15:05:15] I possibly know of a place, depending on what 'mac os server hardware' means [15:05:27] --- stevenjenkins has left [15:05:31] Anything Intel, really. [15:05:38] A mini would do... [15:05:50] --- stevenjenkins has become available [15:06:20] It just needs to be something that's permanently online, and that the owner is happy with us kicking off OpenAFS builds on every 10 minutes or so ... [15:06:40] By "fedora" you would mean the latest stable version? [15:06:51] --- abo has left [15:07:01] Anything we like really. We should be able to virtualise that, so we have more flexibility. [15:07:25] (same place should be able to provide debian and sol10 as well, at least) [15:07:33] --- abo has become available [15:08:06] Excellent. I will get back to you once the infrastructure is built. [15:08:27] something that follows the latest kernel git would be useful [15:08:48] --- jaltman has left: Replaced by new connection [15:08:49] --- jaltman has become available [15:08:54] --- haba has become available [15:09:11] Would we want to verify every change against that, though, or just changes which make it in to master? [15:09:34] we had something like that (against linux git) for tinderbox at once point [15:09:38] one point [15:09:42] We'd also need to be careful to pick up whether it's an OpenAFS change, or a kernel change, which has broken us. [15:10:14] My intention is that buildbot won't replace what Mike did with tinderbox - just that it will be closely coupled to gerrit, so it can build prospective changes, rather than just tracking master. [15:10:21] true - that one would be useful to get a heads up on changes, not to reject a change [15:11:14] We might actually want to do that one the other way around, and trigger it from changes in kernel git, rather than those in openafs. [15:15:28] yes, that's what I thought it was for; so we're notified as early as possible of changes that break us [15:15:35] (linux changes, that is) [15:17:11] Yeh. Marc's doing a grand job of tracking those at the moment - if we can make that easier, it can only be a good thing ... [15:17:14] i rarely run a kernel that's more that a week old, so we've been catching most linux changes pretty early. [15:17:31] but it's manual - automating it would be good [15:17:35] Definitely. [15:18:11] Hopefully the offers of hardware will come through, and we'll have somewhere to run all of this... [15:24:30] --- deason has left [15:25:08] --- deason has become available [15:40:34] --- dev-zero@jabber.org has become available [15:40:55] --- dev-zero@jabber.org has left [15:41:11] --- dev-zero@jabber.org has become available [16:19:00] --- jaltman has left: Disconnected [16:22:13] --- dev-zero@jabber.org has left: Replaced by new connection [16:22:14] --- dev-zero@jabber.org has become available [16:37:25] --- dev-zero@jabber.org has left: Lost connection [16:45:50] --- pod has left [16:56:49] --- dev-zero@jabber.org has become available [17:01:44] --- Russ has left: Disconnected [17:30:31] Well, it looks like I can shave 15s or so from an OpenAFS build by using splice/sendpage to remove a copy from the store path. Almost worth the effort ... [17:36:24] --- Russ has become available [17:40:12] that's 15s out of how many? if I shaved 15s off my build time I'd be into negative numbers... :) [17:41:37] 6minutes on my machine here :( [17:41:43] (that's building in AFS) [17:44:01] Patch is at /afs/inf.ed.ac.uk/user/s/sxw/Public/openafs-splice.patch [17:44:59] (modern kernels want the code path that's currently #if 0'd in that patch. I need to write a configure test ... ) [17:48:05] so I should switch the if 0 with the else part to try this on a new kernel? [17:48:13] Yes. [17:48:43] (sendpage might actually still work on a new kernel, put use the splice_direct stuff is more correct) [17:49:04] just changing the #if 0 to #if 1 should do the trick. [17:51:01] hmm, doesn't apply as is [17:51:55] Oh d'Oh. [17:52:02] There's an earlier patch in that series. [17:53:46] /afs/inf.ed.ac.uk/user/s/sxw/Public/openafs-storeback.patch [17:53:52] needs to be applied first ... [18:01:44] --- dev-zero@jabber.org has left [18:01:56] --- dev-zero@jabber.org has become available [19:26:06] --- Jeffrey Altman has become available [19:31:00] --- mdionne has left [20:46:48] --- dev-zero@jabber.org has left [20:47:20] --- dev-zero@jabber.org has become available [20:59:36] --- Russ has left: Disconnected [21:16:47] --- dev-zero@jabber.org has left [21:16:58] --- dev-zero@jabber.org has become available [22:21:53] --- dev-zero@jabber.org has left [22:22:06] --- dev-zero@jabber.org has become available [22:39:07] --- haba has left: Lost connection [22:39:07] --- abo has left: Lost connection [22:39:50] --- abo has become available [22:41:46] --- haba has become available [23:07:11] --- deason has left