Camelia, the Perl 6 bug

IRC log for #gluster, 2012-12-08

| Channels | #gluster index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:17 stopbit joined #gluster
00:22 TSM2 joined #gluster
00:58 hchiramm_ joined #gluster
01:04 bala1 joined #gluster
01:19 rammses_theold joined #gluster
01:19 hchiramm_ joined #gluster
01:45 mooperd joined #gluster
02:35 crashmag joined #gluster
02:48 hagarth joined #gluster
03:52 bauruine joined #gluster
04:00 yinyin joined #gluster
04:52 robo joined #gluster
05:39 avati joined #gluster
06:06 glusterbot New news from newglusterbugs: [Bug 885281] Swift integration Gluster_DiskFIle.unlink() performs directory listing before deleting file ... scalable? <http://goo.gl/pR86P>
06:34 yinyin joined #gluster
06:45 yinyin joined #gluster
06:52 hchiramm_ joined #gluster
06:56 bala joined #gluster
07:13 hchiramm_ joined #gluster
07:24 yinyin joined #gluster
07:29 efries joined #gluster
07:33 hchiramm_ joined #gluster
07:59 yinyin joined #gluster
08:19 yinyin joined #gluster
08:57 nightwalk joined #gluster
09:36 hchiramm_ joined #gluster
09:43 nightwalk joined #gluster
09:53 yinyin joined #gluster
09:53 yinyin_ joined #gluster
10:14 quillo joined #gluster
10:21 dan_a joined #gluster
10:27 hchiramm_ joined #gluster
10:36 gbrand_ joined #gluster
10:49 mooperd joined #gluster
11:33 TSM2 joined #gluster
11:45 mooperd joined #gluster
12:00 andreask joined #gluster
12:01 andreask left #gluster
14:18 robo joined #gluster
14:21 chirino joined #gluster
15:04 bala joined #gluster
15:28 gbrand_ joined #gluster
15:49 andreask joined #gluster
15:49 andreask left #gluster
16:12 pete__ joined #gluster
16:13 peterlin joined #gluster
16:14 peterlin Hello, I could use a hand: I just connected a handful of additional clients using NFS. Now im getting "nfs: server 1.1.1.1 not responding, still trying". Any hints on what I can do to improve things?
16:15 peterlin If I kill a few clients the nfs-server comes  back: nfs: server 1.1.1.1 OK"
16:47 ron-slc joined #gluster
17:00 andreask joined #gluster
17:02 andreask left #gluster
17:23 JoeJulian peterlin: Check the nfs.log on the server your nfs clients are mounting from for clues .
17:25 gbrand_ joined #gluster
17:26 peterlin joejulian: Good idea. Found this: "0-nfs-NLM: Unable to get NLM port of the client. Is the firewall running on client?"
17:26 peterlin no firewall though, and I think statd is running.
17:27 JoeJulian Not sure why that would block either...
17:29 peterlin portmap is running. There are very many locks though, probably around 300-400 processes contending. Will disable locking and see if everything moves along...
17:50 Alpinist joined #gluster
18:15 designbybeck joined #gluster
18:18 peterlin Can I tune io-threads somehow to get better nfs performance with many clients?
18:21 JoeJulian One sure way to find out...
18:23 JoeJulian Might consider trying client-side io-threads too. I'm not sure if that affects the nfs service, but it would be interesting to find out. set "client-io-threads on"
18:32 peterlin I havent changed any io-threads settings on this volume. What is the default values?
18:33 JoeJulian gluster volume set help
18:38 peterlin The default seems to be 16, that should it it. What would client-io-threads on" do, and is it safe on a live system?
18:39 JoeJulian Sorry if I'm coming across as terse. I'm trying to figure out why my backup server stopped communicating on the network between 01:45:27 and 01:57:34, then again at 05:12:14 and 05:23:42 with absolutely no log entries.
18:39 peterlin the "glusterfs" process that seems to be the nfs server is sitting at around 70% cpu.
18:39 peterlin Bad switch?
18:40 JoeJulian I'm wondering that myself, but none of the other machines on that switch had any problems.
18:41 JoeJulian So the client io-threads... not really sure, but it's been suggested when many threads are accessing the volume through the fuse client. It's safe to turn on and off on a live system.
18:44 stopbit joined #gluster
18:47 peterlin With "nolocks", the nfs client does not seem to do much caching?
18:53 JoeJulian Seems odd since it did with 3.[12] which didn't even have locking. Check /proc/fs/fscache/stats
19:00 peterlin I don't have that file/dir, (debian)
19:03 JoeJulian Huh, I always though that was loaded by default. Try modprobe fscache and/or adding "fsc" to your nfs mount options.
19:04 andreask joined #gluster
19:04 JoeJulian Caveat emptor, I don't use the nfs features of glusterfs.
19:05 TSM2 joined #gluster
19:09 andreask1 joined #gluster
19:11 andreask joined #gluster
19:13 peterlin Found it on a newer install :). What am I looking at, all the numbers say 0
19:14 andreask2 joined #gluster
19:14 andreask joined #gluster
19:22 andreask joined #gluster
19:22 efries joined #gluster
19:29 andreask joined #gluster
19:46 peterlin Somehow files that users upload via FTP gets owned by root. Why could that be?
19:49 andreask left #gluster
19:59 peterlin I need some urgent help: All files that I create using any user gets owned by root, instead of the correct user.
20:00 andreask joined #gluster
20:01 andreask left #gluster
20:02 andreask1 joined #gluster
20:02 andreask1 left #gluster
20:11 peterlin Can this happen when i mount using a 32 bit client?
20:11 noob2 joined #gluster
20:12 noob2 left #gluster
20:14 JoeJulian shouldn't be happening since 2.0.2. If you chown them, does it stick?
20:17 peterlin JoeJulian: Yes it does, as the user I can chown the file, and it works after that.
20:17 peterlin This is on 3.3.0
20:18 JoeJulian Can you upgrade to 3.3.1? I don't know of a specific bug that would cause that, but there were some that caused weird communication issues in 3.3.0.
20:18 peterlin I created a file from a 64 bit nfs client, and that worked?
20:19 JoeJulian There's an nfs.ino32 setting, could try playing with that.
20:20 JoeJulian Maybe just a remount on the 32 bit client?
20:20 JoeJulian Doesnt' seem likely though.
20:21 peterlin Tried turning it on, and rebooted the client, no effect
20:21 JoeJulian This was working before, right?
20:22 peterlin I'm not sure actually. I never thought of it before.
20:23 peterlin The kernel on the client is kind of old, 2.6.26.
20:27 m0zes I've seen weird ownership changes if the client glusterfs dies while it has a file open. sometimes the file will revert to root:root instead of what it was... this was <=glusterfs-3.2.6 iirc. I don't think I've seen it in 3.2.7.
20:31 peterlin Got lots of stale file handles when i turnen nfs.ino32 off again :(
20:33 peterlin For now I will have to revert back to our old platform, we have to rush that through right now :(
20:41 peterlin Does it make sense to the difference between 32 and 64 bit client?
20:47 JoeJulian Not to me, but could you please file a bug report?
20:47 glusterbot http://goo.gl/UUuCq
21:08 peterlin I will have to check more tomorrow, its late here. We just got everything working again with our old nfs server, so I have some time. I would be prepared to pay someone to help me reason this thing out?
21:09 peterlin Btw, i'm using the same cluster for VM hosting. It works very well indeed. Even if i shut a node down, pull a power cord or whatever everything keeps on executing, and after the node comes back online it heals itself well (although it takes 6 hours, but no downtime for VMs)
21:27 mooperd joined #gluster
21:36 robo joined #gluster

| Channels | #gluster index | Today | | Search | Google Search | Plain-Text | summary