#ceph IRC Log

Index

IRC Log for 2012-04-02

Timestamps are in GMT/BST.

[0:23] * buck (~buck@bender.soe.ucsc.edu) Quit (Quit: Leaving)
[0:38] * eroanok (~Adium@cpe-74-73-141-40.nyc.res.rr.com) has joined #ceph
[0:46] * pruby (~tim@leibniz.catalyst.net.nz) has joined #ceph
[1:14] * PaulMachTech (~smuxi@203.206.186.1) has joined #ceph
[1:14] * PaulMachTech (~smuxi@203.206.186.1) Quit (Read error: Connection reset by peer)
[1:15] * BManojlovic (~steki@212.200.240.52) Quit (Quit: Ja odoh a vi sta 'ocete...)
[1:34] * joao (~joao@89.181.151.120) has joined #ceph
[1:34] * joao (~joao@89.181.151.120) Quit ()
[2:37] * yoshi (~yoshi@p1062-ipngn1901marunouchi.tokyo.ocn.ne.jp) has joined #ceph
[3:25] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) Quit (Ping timeout: 480 seconds)
[3:27] * lofejndif (~lsqavnbok@04ZAACE7D.tor-irc.dnsbl.oftc.net) has joined #ceph
[3:27] * lofejndif (~lsqavnbok@04ZAACE7D.tor-irc.dnsbl.oftc.net) Quit (Remote host closed the connection)
[3:33] * lofejndif (~lsqavnbok@28IAADOVT.tor-irc.dnsbl.oftc.net) has joined #ceph
[3:36] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) has joined #ceph
[4:15] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) Quit (Remote host closed the connection)
[4:31] * lofejndif (~lsqavnbok@28IAADOVT.tor-irc.dnsbl.oftc.net) Quit (Quit: Leaving)
[5:07] * lofejndif (~lsqavnbok@659AAA89C.tor-irc.dnsbl.oftc.net) has joined #ceph
[5:48] * lofejndif (~lsqavnbok@659AAA89C.tor-irc.dnsbl.oftc.net) Quit (Quit: Leaving)
[6:23] * lxo (~aoliva@lxo.user.oftc.net) Quit (Read error: Operation timed out)
[6:36] * lxo (~aoliva@lxo.user.oftc.net) has joined #ceph
[6:42] * f4m8_ is now known as f4m8
[6:48] * eroanok (~Adium@cpe-74-73-141-40.nyc.res.rr.com) Quit (Quit: Leaving.)
[6:55] * LarsFronius (~LarsFroni@e182093022.adsl.alicedsl.de) has joined #ceph
[6:55] * adjohn (~adjohn@p1062-ipngn1901marunouchi.tokyo.ocn.ne.jp) has joined #ceph
[7:05] * LarsFronius (~LarsFroni@e182093022.adsl.alicedsl.de) Quit (Quit: LarsFronius)
[8:21] * loicd (~loic@magenta.dachary.org) Quit (Ping timeout: 480 seconds)
[9:21] * loicd (~loic@83.167.43.235) has joined #ceph
[9:23] * loicd (~loic@83.167.43.235) has left #ceph
[9:25] * andreask (~andreas@chello062178013131.5.11.vie.surfer.at) has joined #ceph
[10:22] * Qten (~Q@ppp59-167-157-24.static.internode.on.net) has joined #ceph
[10:38] * adjohn (~adjohn@p1062-ipngn1901marunouchi.tokyo.ocn.ne.jp) Quit (Quit: adjohn)
[11:06] * adjohn (~adjohn@p1062-ipngn1901marunouchi.tokyo.ocn.ne.jp) has joined #ceph
[11:26] * yoshi (~yoshi@p1062-ipngn1901marunouchi.tokyo.ocn.ne.jp) Quit (Remote host closed the connection)
[11:32] * adjohn (~adjohn@p1062-ipngn1901marunouchi.tokyo.ocn.ne.jp) Quit (Quit: adjohn)
[11:50] * morse (~morse@supercomputing.univpm.it) has joined #ceph
[12:14] <Qten> heya's i noticed 44.1 is out are we out of beta yet? 8)
[12:16] * LarsFronius (~LarsFroni@manaslu.utelisys.net) has joined #ceph
[12:26] <chaos_> hello ;-) I've not so small question, we want to add 2 more osd (now we have 2), because we are running out of space (5T isn't enough ;(), our crush map looks like this http://wklej.org/hash/45ac6b84b9/, what will happen if I'll just add 2 osd? every data will be replicated at each node? And just after adding node data will be synchronized, which means lots of unwanted traffic? I want to achive something like raid 0+1 (one old node + old new node mirrored
[12:41] * nhorman (~nhorman@99-127-245-201.lightspeed.rlghnc.sbcglobal.net) has joined #ceph
[13:06] * joao (~JL@89-181-151-120.net.novis.pt) has joined #ceph
[13:11] * LarsFronius (~LarsFroni@manaslu.utelisys.net) Quit (Quit: LarsFronius)
[13:16] * loicd (~loic@83.167.43.235) has joined #ceph
[13:48] <NaioN> chaos_: see the page in the wiki for adding an osd
[13:48] * LarsFronius (~LarsFroni@manaslu.utelisys.net) has joined #ceph
[13:48] <NaioN> you can edit the crushmap manually
[13:49] <NaioN> chaos_: http://ceph.newdream.net/wiki/OSD_cluster_expansion/contraction
[13:50] <NaioN> the last time I used it, the osd got added to the crushman automaticly, but that's not always
[13:50] * BManojlovic (~steki@91.195.39.5) has joined #ceph
[13:52] <NaioN> but normally if you add an osd and it gets included in the crushmap some data will change from osd
[13:52] <NaioN> it depends on the crushmap how much data and to which osd it gets moved
[14:03] * mtk (~mtk@ool-44c35967.dyn.optonline.net) has joined #ceph
[14:11] <darkfader> hehe last time i tried to change my crushmap i blew up everything. maybe it was just rebalancing too heavily. but it was fun
[14:30] <loicd> Hi, I'm trying to find professional support for RBD . The idea is to fix problems (because it's still young) as quickly as possible.
[14:31] * LarsFronius (~LarsFroni@manaslu.utelisys.net) Quit (Quit: LarsFronius)
[14:56] * eroanok (~Adium@cpe-74-73-141-40.nyc.res.rr.com) has joined #ceph
[14:57] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) has joined #ceph
[14:59] <joao> loicd, you should contact info@ceph.com
[15:00] <joao> they surely will give you all the informations you need :)
[15:01] * eroanok (~Adium@cpe-74-73-141-40.nyc.res.rr.com) Quit (Quit: Leaving.)
[15:02] <loicd> joao: I did last week but no reply yet. Hence my 'live chat' question ;-)
[15:03] <joao> I see
[15:04] <joao> maybe you should wait a couple more hours; later on the afternoon the team should be available on the channel
[15:04] <joao> later on *my* afternoon, I mean
[15:04] <joao> :)
[15:05] <joao> they should start showing up in a couple of hours
[15:10] <loicd> :-)
[15:16] <chaos_> NaioN, i know how to add osd;p but I don't have idea how crush map should like like for raid0+1 wannabe;p
[15:21] * LarsFronius (~LarsFroni@manaslu.utelisys.net) has joined #ceph
[15:39] <NaioN> chaos_: aha but you don't have a RAID0+1
[15:40] <NaioN> data gets distributed over the different osd's so read and writes to different objects gets distributed over the different osd's
[15:41] <NaioN> and if you have a replication level the objects get replicated one or more times
[15:41] <NaioN> but it isn't that you read or write to different replications at the same time for the same object
[15:42] <NaioN> every object has a first/master and that one gets replicated
[15:42] <NaioN> so you only read or write to the first/master object
[15:42] * f4m8 is now known as f4m8_
[15:43] * pmcarthur (~peter@nyv-exweb.iac.com) Quit (Remote host closed the connection)
[15:43] <NaioN> do you use rbd's or cephfs?
[15:44] <chaos_> cephfs
[15:48] <NaioN> aha with cephfs every file is a object if i'm correct
[15:48] <NaioN> till 4MB (default) and larger files get split into multiple objects
[15:49] <NaioN> if i'm correct
[15:50] <NaioN> but for smaller files you read or write to one osd
[15:50] * eroanok (~Adium@206.217.92.10) has joined #ceph
[15:50] <NaioN> but if you read/write to a lot of files at the same time you get a nice even distribution of load on all the osds
[15:50] <chaos_> hmm
[15:51] * eroanok (~Adium@206.217.92.10) has left #ceph
[15:52] <NaioN> so if you read/write to one file all the time you're doing all the io to one osd.... that osd will replicate the file/object in the background to 1 or more different osd's (depending on your replication level)
[15:56] * nyeates (~nyeates@pool-173-59-237-75.bltmmd.fios.verizon.net) has joined #ceph
[15:59] * rick1 (~rick@659AAA9RK.tor-irc.dnsbl.oftc.net) has joined #ceph
[16:01] <- *rick1* hi
[16:01] <rick1> hi
[16:02] <rick1> hello there
[16:15] <Dieter_be> hi
[16:23] * rick1 (~rick@659AAA9RK.tor-irc.dnsbl.oftc.net) has left #ceph
[17:07] * LarsFronius (~LarsFroni@manaslu.utelisys.net) Quit (Quit: LarsFronius)
[17:21] * LarsFronius (~LarsFroni@manaslu.utelisys.net) has joined #ceph
[17:34] * oliver1 (~oliver@p4FFFE518.dip.t-dialin.net) has joined #ceph
[17:44] * LarsFronius (~LarsFroni@manaslu.utelisys.net) Quit (Quit: LarsFronius)
[17:48] * BManojlovic (~steki@91.195.39.5) Quit (Quit: Ja odoh a vi sta 'ocete...)
[17:52] * bchrisman (~Adium@c-76-103-130-94.hsd1.ca.comcast.net) Quit (Quit: Leaving.)
[17:57] * adjohn (~adjohn@s24.GtokyoFL16.vectant.ne.jp) has joined #ceph
[17:57] * adjohn (~adjohn@s24.GtokyoFL16.vectant.ne.jp) Quit ()
[17:57] * LarsFronius (~LarsFroni@manaslu.utelisys.net) has joined #ceph
[18:01] * LarsFronius_ (~LarsFroni@manaslu.utelisys.net) has joined #ceph
[18:01] * LarsFronius (~LarsFroni@manaslu.utelisys.net) Quit (Read error: Connection reset by peer)
[18:01] * LarsFronius_ is now known as LarsFronius
[18:03] * LarsFronius (~LarsFroni@manaslu.utelisys.net) Quit ()
[18:08] * oliver1 (~oliver@p4FFFE518.dip.t-dialin.net) has left #ceph
[18:10] * Tv|work (~Tv_@aon.hq.newdream.net) has joined #ceph
[18:13] * jmlowe (~Adium@129-79-195-139.dhcp-bl.indiana.edu) has joined #ceph
[18:14] <jmlowe> Would somebody mind poking http://ceph.newdream.net with sharp stick, it seems to be unresponsive?
[18:15] * sagelap (~sage@50-1-53-2.dedicated.static.sonic.net) has joined #ceph
[18:16] <jmlowe> works now, thanks if you did something
[18:18] * __jt__ (~james@jamestaylor.org) Quit (Quit: leaving)
[18:18] * __jt__ (~james@jamestaylor.org) has joined #ceph
[18:20] * joshd (~joshd@aon.hq.newdream.net) has joined #ceph
[18:39] <chaos_> Hello newdream ;-)
[18:39] <chaos_> maybe you will help you
[18:40] <Tv|work> good morning
[18:40] <Tv|work> i always try to help me
[18:40] <chaos_> "we want to add 2 more osd (now we have 2), because we are running out of space (5T isn't enough ;(), our crush map looks like this http://wklej.org/hash/45ac6b84b9/, what will happen if I'll just add 2 osd? every data will be replicated at each node? And just after adding node data will be synchronized, which means lots of unwanted traffic? I want to achive something like raid 0+1 (one old node + old new node mirrored to another group one old + one new).
[18:40] <chaos_> Tv|work, I know i can count on you ;-)
[18:42] <Tv|work> chaos_: http://ceph.newdream.net/docs/master/ops/manage/grow/osd/ is the docs for this, i think you shouldn't worry too much about the unwanted traffic
[18:43] <Tv|work> chaos_: if you need to handle new and old nodes specially, you need to create that structure in your crush map; but most likely you don't need to
[18:46] <chaos_> Tv|work, what about weight? should I do something like o1 - 0.5, n1 - 0.5, o2 - 0.5, n2 - 0.5? (1st old node - o1, 2nd old - o2, 1st new - n1, 2nd new - n2)
[18:46] <Tv|work> chaos_: if you new nodes have more storage or are faster, then yes
[18:46] <Tv|work> chaos_: or you can use weights to "ease in" to the new layout
[18:46] <chaos_> they will be the same as old
[18:46] <sagelap> joshd: i pushed a wip-discard branch which adds librados zero calls. in retrospect, though, it might be better to use ObjectWriteOperation for everything and try to minimize (instead of expand) the API...
[18:46] <Tv|work> setting all weights to 0.5 has no effect; it's all relative
[18:47] <chaos_> oh
[18:47] <joshd> sagelap: looking
[18:48] <chaos_> so, if I just add two new nodes, everything will be as I want (mirror and replication)? automagicaly?:P
[18:49] <chaos_> erm
[18:49] <chaos_> mirror and replication is the same;p don't read this
[18:49] <chaos_> i mean expanded space and mirroring
[18:50] <Tv|work> chaos_: with the default crush map, ceph will spread the stored objects across all osds available
[18:50] <sagelap> joshd: it would actually be satisfying to rip out aio_* (except for aio_operate() etc.) and force everyone into a single paradigm
[18:52] <joshd> sagelap: it's nice to have the sync wrappers for e.g. glance, but we could probably unify the implementations of aio/sync versions more
[18:53] <chaos_> Tv|work, but under rule data i see type replicated
[18:53] <Tv|work> chaos_: and this worries you because...?
[18:54] <chaos_> data is replicated, yes?:P
[18:54] <Tv|work> yes
[18:54] <chaos_> when i have 2 osd, one can be down
[18:54] <chaos_> and data is accessible
[18:55] <Tv|work> the count of replicas is controlled by "ceph osd pool set POOL size NUM", http://ceph.newdream.net/docs/master/control/#osd-subsystem
[18:59] <chaos_> ok, now get it
[18:59] <chaos_> I'll try to expand our cluster tomorrow ;-) Hopefully without any disaster.
[19:00] <sagelap> joshd: yeah. so, i'm inclned not to add those librados aio calls, then, and use the *Operation methods. that code is wrapped up in the callback stuff in librbd, though, which is in flux in your branch...
[19:00] <chaos_> thx Tv|work for clearing out some things ;-)
[19:01] <sagelap> so this should probably wait for those commits to settle down :/
[19:01] * perplexed (~ncampbell@216.113.168.141) has joined #ceph
[19:02] <joshd> sagelap: they can be rebased without too much trouble
[19:03] <joshd> sagelap: I like the ObjectOperation idea, but we don't have a C interface for it
[19:03] <sagelap> hmm
[19:04] <sagelap> that's fine for librbd. i guess when we want C aio zero that's when the ObjectOperation stuff needs to be added to that api
[19:05] <nhm> joshd: how goes the latency work?
[19:05] <nhm> er, sorry, that was for same
[19:05] * bchrisman (~Adium@108.60.121.114) has joined #ceph
[19:05] <nhm> SAM rather.
[19:05] * nhm needs coffee
[19:06] <chaos_> what is default pool name? or how i can list pools
[19:06] <joshd> sagelap: I guess that's ok for now
[19:08] * yehudasa (~yehudasa@aon.hq.newdream.net) Quit (Remote host closed the connection)
[19:08] <joshd> sagelap: I think there are good ways to reduce the duplicate code in IoCtxImpl since the methods follow the same couple patterns, but that's an idea for another day
[19:08] <Tv|work> chaos_: default install has pools data, metadata, rbd
[19:09] <sagelap> joshd: yeah
[19:09] <Tv|work> chaos_: http://ceph.newdream.net/docs/master/man/8/rados/
[19:09] * chutzpah (~chutz@216.174.109.254) has joined #ceph
[19:10] * yehudasa (~yehudasa@aon.hq.newdream.net) has joined #ceph
[19:10] <chaos_> Tv|work, you have doc link for everything?:>
[19:10] <Tv|work> chaos_: not yet ;)
[19:12] <Tv|work> or maybe i should say... not yet: http://ceph.newdream.net/docs/master/search/?q=todo&check_keywords=yes&area=default
[19:14] * tjikkun (~tjikkun@82-169-255-84.ip.telfort.nl) Quit (Remote host closed the connection)
[19:24] * tjikkun (~tjikkun@82-169-255-84.ip.telfort.nl) has joined #ceph
[19:26] <nhm> kaboom
[19:27] * Oliver1 (~oliver1@ip-37-24-160-195.unitymediagroup.de) has joined #ceph
[19:31] <Tv|work> nhm: new game: vidyo pictionary
[19:31] <nhm> Tv|work: I'll send you guys a bottle of tequilla and every time my client segfaults you have to take a shot.
[19:32] <Tv|work> nhm: more like, every time vidyo fails corp it does *not* get a drink
[19:32] <Tv|work> send negative shots
[19:32] <nhm> excellent
[19:32] <nhm> Nothing like a little motivation...
[19:35] <Tv|work> oh and for the support watch, the one thing really remaining from last week is the "leveldb does not respect our cflags" thread with Jonathan Dieter.. yehudasa sjust
[19:35] <Tv|work> i did not see any resolution and didn't have enough to put into it
[19:36] <Tv|work> +time
[19:42] * morse (~morse@supercomputing.univpm.it) Quit (Remote host closed the connection)
[19:50] * sagelap (~sage@50-1-53-2.dedicated.static.sonic.net) Quit (Ping timeout: 480 seconds)
[19:52] * loicd (~loic@83.167.43.235) Quit (Quit: Leaving.)
[19:52] * dmick (~dmick@aon.hq.newdream.net) has joined #ceph
[20:03] * BManojlovic (~steki@212.200.240.52) has joined #ceph
[20:23] * verwilst (~verwilst@dD5769628.access.telenet.be) has joined #ceph
[20:25] * sagelap (~sage@50-1-53-2.dedicated.static.sonic.net) has joined #ceph
[20:33] * loicd (~loic@magenta.dachary.org) has joined #ceph
[20:40] * sagelap (~sage@50-1-53-2.dedicated.static.sonic.net) Quit (Ping timeout: 480 seconds)
[20:43] * mtk (~mtk@ool-44c35967.dyn.optonline.net) Quit (Remote host closed the connection)
[20:58] * mtk (~mtk@ool-44c35967.dyn.optonline.net) has joined #ceph
[21:17] * mtk (~mtk@ool-44c35967.dyn.optonline.net) Quit (Remote host closed the connection)
[21:18] * mtk (~mtk@ool-44c35967.dyn.optonline.net) has joined #ceph
[21:20] * mtk (~mtk@ool-44c35967.dyn.optonline.net) Quit (Remote host closed the connection)
[21:25] * mtk (~mtk@ool-44c35967.dyn.optonline.net) has joined #ceph
[21:46] <Rankin> I'm going to do a small test of ceph, is it possible to set up ceph inside some openvz containers? Or should i create some kvm instances?
[21:47] <Tv|work> Rankin: kvm definitely yes, openvz should work too
[21:48] <Tv|work> i don't think anyone here uses openvz regularly, but i don't see why it wouldn't work
[21:50] <Rankin> Ok, ill give it a shot.
[21:53] <Rankin> I'll need a minimum of two machines to set up a minimal cluster with replication, am I correct?
[21:54] <gregaf> Rankin: yep
[21:55] <gregaf> or two things that look like hard drives, rather
[21:55] <gregaf> they could live in one machine as far as Ceph is concerned; you'll have to decide if that's appropriate for your testing
[21:57] <Rankin> Hmm, think I'm going to use a couple of kvm instances instead as openvz don't have blockdevices, and I don't want to test with loopback devices...
[22:00] <Tv|work> Rankin: if you ignore performance, all ceph wants is subdirectories
[22:02] <gregaf> s/wants/needs
[22:07] * __jt___ (~james@jamestaylor.org) has joined #ceph
[22:07] * __jt___ (~james@jamestaylor.org) Quit ()
[22:18] * perplexed (~ncampbell@216.113.168.141) has left #ceph
[22:22] * jmlowe (~Adium@129-79-195-139.dhcp-bl.indiana.edu) Quit (Quit: Leaving.)
[22:22] * jksM (jks@193.189.93.254) has joined #ceph
[22:23] * nhorman (~nhorman@99-127-245-201.lightspeed.rlghnc.sbcglobal.net) Quit (Quit: Leaving)
[22:25] * ferai (~quassel@quassel.jefferai.org) has joined #ceph
[22:27] * johnl_ (~johnl@2a02:1348:14c:1720:24:19ff:fef0:5c82) has joined #ceph
[22:28] * verwilst (~verwilst@dD5769628.access.telenet.be) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * andreask (~andreas@chello062178013131.5.11.vie.surfer.at) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * pruby (~tim@leibniz.catalyst.net.nz) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * Rankin (~sel@90-228-233-162-no223.tbcn.telia.com) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * wido (~wido@rockbox.widodh.nl) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * jks2 (jks@193.189.93.254) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * jefferai (~quassel@quassel.jefferai.org) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * jantje (~jan@paranoid.nl) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * johnl (~johnl@2a02:1348:14c:1720:24:19ff:fef0:5c82) Quit (reticulum.oftc.net solenoid.oftc.net)
[22:28] * wido (~wido@rockbox.widodh.nl) has joined #ceph
[22:29] * jantje (~jan@paranoid.nl) has joined #ceph
[22:30] * pruby (~tim@leibniz.catalyst.net.nz) has joined #ceph
[22:31] * Rankin (~sel@90-228-233-162-no223.tbcn.telia.com) has joined #ceph
[22:31] * verwilst (~verwilst@dD5769628.access.telenet.be) has joined #ceph
[22:32] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) Quit (Quit: Ex-Chat)
[22:32] * andreask (~andreas@chello062178013131.5.11.vie.surfer.at) has joined #ceph
[22:32] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) has joined #ceph
[22:35] * imjustmatthew (~imjustmat@pool-71-176-223-2.rcmdva.fios.verizon.net) has joined #ceph
[22:48] * bchrisman (~Adium@108.60.121.114) Quit (Quit: Leaving.)
[22:48] * joshd (~joshd@aon.hq.newdream.net) Quit (Ping timeout: 480 seconds)
[22:56] * joshd (~joshd@aon.hq.newdream.net) has joined #ceph
[23:02] * xarthisius (~xarth@hum.astri.uni.torun.pl) has left #ceph
[23:12] <iggy> do you need an mds running if you aren't mounting the cluster (i.e. just rbd)
[23:12] <gregaf> nope
[23:15] * perplexed (~ncampbell@216.113.168.141) has joined #ceph
[23:17] <perplexed> Quick question... how soon after a rados rmpool command is the disk space associated with that pool released? I'm assuming it's a background purge (?)
[23:21] * verwilst (~verwilst@dD5769628.access.telenet.be) Quit (Quit: Ex-Chat)
[23:21] <gregaf> perplexed: varies ?????it is a background purge and I don't know how strictly it's bounded; sjust could tell you better
[23:22] <sjust> perplexed: each pg for that pool is queued for removal when the osd sees the map removing the pool iirc
[23:22] <sjust> realistically, it should happen within minutes, but I don't believe there is a hard bound
[23:24] * sagelap (~sage@50-1-53-18.dedicated.static.sonic.net) has joined #ceph
[23:24] <nhm> sjust: would throughput/latency testing of 0.43 and 0.44 in addition to master be useful?
[23:25] <sjust> 0.44.<1 or whatever DHO is running>
[23:25] <sjust> would be good
[23:30] <perplexed> thx. Also, for a rados bench command.... I have a 40 osd cluster (4 servers, 10 OSD per, each aligned with its own HD).... does the rados bench accurately simulate client connections to all OSD's/servers? Is the benchmark distributing requests across the cluster (v the "ceph osd tell 0 bench" test that seems directed to a specific osd - 0 in that example)?
[23:31] <gregaf> yes, it should distribute requests across the cluster
[23:31] <gregaf> it's conceivable that something's broken (rados bench started out as a dirty hack) but it would surprise m
[23:31] <gregaf> *me
[23:31] <nhm> perplexed: there was a bug in rados bench that was causing it to only hit 1 OSD, but other than that it should be distributing writes across the cluster.
[23:32] * lofejndif (~lsqavnbok@09GAAEJ2A.tor-irc.dnsbl.oftc.net) has joined #ceph
[23:32] <perplexed> thx! I'm on 0.43 currently. Not sure if the bug was fixed in this version?
[23:32] <nhm> perplexed: I think that got fixed a week or two ago.
[23:32] <gregaf> oh right???.I forgot about that recent one
[23:33] <nhm> perplexed: 0.43 was released before the bug was fixed...
[23:34] <perplexed> Issues getting 0.44 to compile on CentOS 6.1, so in interests of time had to fall back to 0.43. If the fix is easily ported back to 0.43's rados I could perhaps apply and re-compile (?) If needed... thx in advance, have to step away from the computer for 30 mins... Will check back later.
[23:34] <nhm> perplexed: The bug is dependent on the length of the hostname, so you may not be hitting it...
[23:35] <gregaf> but the fix is easily applied!
[23:35] <perplexed> fully qualified? I have loooong FQDN's in the env. Hostname portion is 13 chars though. FQDN string length is 34
[23:36] <perplexed> brb
[23:37] <nhm> perplexed: here's the bug: http://tracker.newdream.net/issues/2196
[23:37] <gregaf> just cherry-pick 2ec8f27f58adca40d125051a23547b639ee7d5f6 if you think you're hitting it
[23:38] <nhm> perplexed: it's entirely possible too though that you are hitting legitimate performance issues too...
[23:55] <perplexed> thx. I'll check out the bug.

These logs were automatically created by CephLogBot on irc.oftc.net using the Java IRC LogBot.