#ceph IRC Log

Index

IRC Log for 2010-11-20

Timestamps are in GMT/BST.

[2:23] * cmccabe (~cmccabe@dsl081-243-128.sfo1.dsl.speakeasy.net) Quit (Quit: Leaving.)
[2:25] <jantje> hi
[2:25] * joshd (~joshd@ip-66-33-206-8.dreamhost.com) Quit (Quit: Leaving.)
[2:28] <jantje> sagewk: I can't check my work email right now
[2:28] <jantje> but I just looked at the bug list, and http://tracker.newdream.net/issues/531 is probably identical to what I reported in that email
[3:09] <jantje> sagewk: let me know if you need anything else
[3:09] <jantje> Nite!
[6:33] * lidongyang (~lidongyan@222.126.194.154) Quit (Remote host closed the connection)
[6:34] * lidongyang (~lidongyan@222.126.194.154) has joined #ceph
[10:18] * Yoric (~David@dau94-10-88-189-211-192.fbx.proxad.net) has joined #ceph
[11:04] * lidongyang (~lidongyan@222.126.194.154) Quit (Remote host closed the connection)
[11:06] * lidongyang (~lidongyan@222.126.194.154) has joined #ceph
[11:28] <jantje> tralala
[12:10] * allsystemsarego (~allsystem@188.26.33.21) has joined #ceph
[12:14] * sentinel_e86 (~sentinel_@188.226.51.71) Quit (Quit: sh** happened)
[12:15] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:17] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:18] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:21] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:22] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:24] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:25] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:28] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:29] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:31] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:32] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:35] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:36] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:38] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:40] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:42] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:43] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:45] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:47] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:49] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:50] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:52] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:54] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[12:56] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[12:57] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:00] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[13:01] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:03] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[13:04] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:07] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[13:08] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:10] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[13:11] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:14] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[13:15] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:17] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[13:18] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:21] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[13:22] * sentinel_e86 (~sentinel_@188.226.51.71) has joined #ceph
[13:24] * sentinel_e86 (~sentinel_@188.226.51.71) Quit ()
[14:19] <wido> jantje: shouldn't you take a weekend? ;)
[14:31] * tjikkun (~tjikkun@2001:7b8:356:0:204:bff:fe80:8080) Quit (Ping timeout: 480 seconds)
[14:31] * tjikkun (~tjikkun@2001:7b8:356:0:204:bff:fe80:8080) has joined #ceph
[14:36] * Yoric (~David@dau94-10-88-189-211-192.fbx.proxad.net) Quit (Quit: Yoric)
[14:46] <jantje> wido: well, yea, but ... it's so addictive
[15:24] <wido> jantje: indeed :)
[15:26] <jantje> i just had this weird thing with iozone
[15:26] <jantje> my random writes were faster than my random reads
[15:26] <jantje> iostat on the osd servers showed very low usage
[15:26] <jantje> still have to investigate some more
[15:27] <jantje> but it's just weird, but there probably is an explanation! :)
[15:41] <wido> jantje: I
[15:41] <wido> I'm seeing lower reads then writes too
[15:42] <wido> When my cluster was optimal, I got 150MB/sec writes, but about 60MB/sec read
[15:45] <jantje> random read/write ?
[15:45] <jantje> that pretty good :-)
[15:45] <jantje> well, you probably have lots of OSDs
[15:46] <jantje> 13:47 < jantje> Children see throughput for 10 random readers = 1997.05 KB/sec with 6 OSD's, which comes down to 500KB/sec random read for each disk
[15:46] <jantje> 13:47 < jantje> Children see throughput for 10 random writers = 6493.89 KB/sec
[15:47] <jantje> I should go and see how fast it is on a local disk and compare it to a single OSD
[15:47] <jantje> en then do the math for multiple OSD's
[15:59] <wido> jantje: that was with 32 OSD's
[15:59] <wido> but lacking good journaling, I've got the ok for 2011 to buy a brand new cluster to do better performance testing
[15:59] <wido> SSD's for journaling, etc
[16:00] * tjikkun (~tjikkun@2001:7b8:356:0:204:bff:fe80:8080) Quit (Ping timeout: 480 seconds)
[16:24] <jantje> nice :)
[16:24] <jantje> i just do journaling on a memory fs .. for now
[16:25] <jantje> ocz vertex 2 are pretty fast
[16:25] <jantje> 50k iops
[16:25] <jantje> and affordable
[16:25] <jantje> I think that would be my choice to host the journals
[16:29] <wido> jantje: indeed
[16:29] <wido> btw, from which belgium part are you? antwerp?
[16:29] <jantje> yea
[16:29] <jantje> very close to
[16:30] <jantje> (20km to the south)
[16:30] <wido> we kunnen ook nederlands hier praten, snapt de rest weinig van...
[16:30] <jantje> dat is waar :)
[16:30] <jantje> but I don't want to get kicked out :P
[16:30] <jantje> nuja, het maakt eigenlijk niet uit
[16:30] <jantje> ze moeten maar een vertaler gebruiken :)
[16:30] <wido> nee, zolang er niemand anders is
[16:31] <wido> de devs slapen nu toch
[16:32] <jantje> ik probeer ceph te gebruiken om een distributed build storage op te zetten
[16:32] <jantje> maar met 6 OSD's en memory journaling is het trager als een single disk
[16:32] <jantje> heb ook al bonding gebruikt enzo
[16:32] <jantje> te veel delay gok ik
[16:32] <jantje> of misschien de random read/writes dat problemen geven
[16:33] <jantje> weet niet goed hoe ik de workload moet identificieren
[16:33] <wido> ik heb gemerkt dat een SSD voor journaling goed werkt, de I/O paden van de data en journal scheiden scheelt heel veel
[16:34] <wido> vergeet ook niet, hoe hoger je replication level, hoe langer een write duurt
[16:34] <jantje> maar de journal files op een momory filesystem zetten is toch zowiso sneller, net?
[16:34] <jantje> ja heb default, 2x ds
[16:34] <jantje> *dus
[16:34] <wido> ja, zeker, alleen bij een reboot ben je je memory fs kwijt
[16:34] <jantje> true
[16:34] <jantje> ik begin toch elke keer met een empty filesystem
[16:34] <wido> maar voor de test moet het prima werken
[16:35] <wido> benchmark je ook met radosbench?
[16:35] <jantje> nope
[16:35] <jantje> weet er nie zo veel vanaf
[16:36] <wido> kan je ook gebruiken om eens goed te kijken wat er gebeurd
[16:36] <wido> maar zie ook: http://tracker.newdream.net/issues/538
[16:36] <wido> en als je de OSD's individueel benchmarked ( http://ceph.newdream.net/wiki/Troubleshooting ) geven ze dan wel een goede performance
[16:36] <wido> ?
[16:36] <wido> brb
[16:41] <jantje> ja
[16:41] <jantje> maar zijn sequential writes, niet?
[16:41] <jantje> random read/write + delay is het grootste probleem denk ik
[16:41] * jantje moet ook even weg
[16:43] <wido> jantje: Ja, maar de writes gaan direct naar de OSD
[16:44] <wido> daar heb je netwerk delay, maar die heb je ook met NFS of iets dergelijks
[16:48] <wido> ik ben afk voor vandaag, succes nog! Als je nog vragen hebt, shoot!
[16:48] <wido> (End of conversation in Dutch ;) )
[16:51] <jantje> bye :)
[17:22] * lidongyang (~lidongyan@222.126.194.154) Quit (Remote host closed the connection)
[17:22] * lidongyang (~lidongyan@222.126.194.154) has joined #ceph
[17:28] * tjikkun (~tjikkun@2001:7b8:356:0:204:bff:fe80:8080) has joined #ceph
[17:53] * greglap (~Adium@166.205.139.168) has joined #ceph
[18:21] * allsystemsarego_ (~allsystem@188.26.32.15) has joined #ceph
[18:28] * allsystemsarego (~allsystem@188.26.33.21) Quit (Ping timeout: 480 seconds)
[19:00] * greglap (~Adium@166.205.139.168) Quit (Read error: Connection reset by peer)
[19:13] * greglap (~Adium@cpe-76-90-74-194.socal.res.rr.com) has joined #ceph
[19:30] * Yoric (~David@dau94-10-88-189-211-192.fbx.proxad.net) has joined #ceph
[21:26] * allsystemsarego_ (~allsystem@188.26.32.15) Quit (Quit: Leaving)

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