Camelia, the Perl 6 bug

IRC log for #gluster-dev, 2013-01-31

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

All times shown according to UTC.

Time Nick Message
03:04 overclk joined #gluster-dev
03:05 bharata joined #gluster-dev
03:47 hagarth joined #gluster-dev
03:52 bulde joined #gluster-dev
04:05 sripathi joined #gluster-dev
04:22 sahina joined #gluster-dev
04:33 pai joined #gluster-dev
04:41 _benoit_ joined #gluster-dev
04:45 vpshastry joined #gluster-dev
04:52 krishnan_p joined #gluster-dev
04:52 krishnan_p #gluster
04:54 _benoit_ joined #gluster-dev
05:32 mohankumar joined #gluster-dev
05:39 raghu joined #gluster-dev
06:06 mohankumar joined #gluster-dev
06:10 deepakcs joined #gluster-dev
06:25 sripathi joined #gluster-dev
06:33 krishnan_p joined #gluster-dev
06:55 bala joined #gluster-dev
06:57 sahina joined #gluster-dev
07:04 krishnan_p joined #gluster-dev
07:05 sripathi joined #gluster-dev
07:29 jbrooks joined #gluster-dev
07:58 sripathi1 joined #gluster-dev
08:02 sripathi joined #gluster-dev
08:27 bulde joined #gluster-dev
08:29 bala joined #gluster-dev
08:46 sripathi joined #gluster-dev
09:05 sahina joined #gluster-dev
09:42 blues-man joined #gluster-dev
10:09 xavih joined #gluster-dev
10:53 vpshastry joined #gluster-dev
11:21 edward joined #gluster-dev
11:49 vpshastry joined #gluster-dev
12:22 kkeithley1 joined #gluster-dev
12:58 jbrooks joined #gluster-dev
13:30 bfoster joined #gluster-dev
13:36 _Bryan_ joined #gluster-dev
13:57 badone joined #gluster-dev
14:04 vpshastry joined #gluster-dev
14:26 puebele1 joined #gluster-dev
14:38 spn joined #gluster-dev
14:40 hagarth joined #gluster-dev
14:56 puebele joined #gluster-dev
15:01 puebele1 joined #gluster-dev
15:09 vpshastry joined #gluster-dev
15:12 edward left #gluster-dev
15:12 puebele joined #gluster-dev
15:13 edward joined #gluster-dev
15:18 puebele joined #gluster-dev
15:30 wushudoin joined #gluster-dev
15:46 puebele1 joined #gluster-dev
15:54 puebele1 joined #gluster-dev
16:01 puebele joined #gluster-dev
16:13 _benoit_ Hi
16:14 _benoit_ Would something like this (https://lists.nongnu.org/archive/htm​l/qemu-devel/2013-01/msg03220.html) be wellcome in the gluster codebase if someone write it ?
16:15 _Bryan_ joined #gluster-dev
16:16 _benoit_ The idea is to mirror all writes and do a vote after doing multiple read of the same block to always get the correct data
16:17 _benoit_ I wrote it for the qemu block layer and I am stuck because I need another feature to be completed before it can be merged
16:17 _benoit_ So I am thinking of rewriting it for glusterfs
16:43 johnmark _benoit_: interesteing
16:44 johnmark _benoit_: I would propose that on gluster-devel
16:44 johnmark but yes, I think it would be interesting
16:45 johnmark but need to make sure it won't clobber our existing quorum features
16:45 johnmark although it doesn't sound like it will
16:46 _benoit_ johnmark: is there any existing translator where it could live ?
16:46 hagarth _benoit_: would you maintain the checksum of a block somewhere?
16:47 _benoit_ hagarth: the idea is to read each copies of the data and take the version which has the majority
16:47 _benoit_ hagarth: and define a threshold for this majority
16:48 _benoit_ hagarth: if no version of the data has more occurence than the threshold it would return -EIO
16:49 _benoit_ hagarth: else it just take the most represented version so bitflip and corruptions are repaired
16:49 hagarth _benoit_: interesting, in case of gluster's afr (replicate) translator, the read is directed to only one of N servers
16:50 hagarth _benoit_: there are mechanisms evolving in afr like outcast to prevent bad reads from reaching the application
16:51 _benoit_ hagarth: what is outcast ?
16:51 hagarth _benoit_: we plan to implement bitrot detection on the server (brick) side
16:52 _benoit_ hagarth: interesting
16:52 hagarth _benoit_:  outcast is a way to label a replica copy as "bad" .. so that no further operations like read are served from that copy
16:53 hagarth _benoit_: the label would be removed after a self-heal makes all replicas consistent
16:54 _benoit_ hagarth: do you think a read quorum can trigger the outcast in afr ?
16:57 _benoit_ hagarth: generally speaking I could devote coding time to implement read quorum in gluster but only if I know that the feature is wellcome and make sense to gluster developpers
16:57 hagarth _benoit_: by default, the read is sent to only one of the replicas (as making multiple network calls for read could be expensive)
16:58 _benoit_ hagarth: is modifying afr do do multiple read reasonable or is it taboo ?
16:59 hagarth _benoit_: i think it might be a good idea to open a thread on gluster-devel
16:59 _benoit_ ok
16:59 hagarth _benoit_: as long as it is optional, it should be ok
16:59 _benoit_ hagarth: I'll open a thread thanks
17:00 hagarth _benoit_: consistency and performance normally don't go together - so it might be a trade off somebody is willing to take
17:00 hagarth _benoit_: so i would not consider it taboo
17:01 johnmark hagarth: +1
17:01 jdarcy Sounds a lot like the Dynamo consistency model.
17:02 * johnmark googles 'Dynamo'
17:02 hagarth jdarcy: are you back?
17:02 jdarcy hagarth: Back?  I never left.
17:02 _benoit_ jdarcy: yes the idea comes from dynamo
17:03 hagarth jdarcy: somehow I was under the impression that you were in lca :)
17:03 jdarcy johnmark: Amazon's original key/value store, providing the basis for Cassandra, Riak, Voldemort, and others.
17:03 jdarcy hagarth: Nope.  I've been at home all week.  John Mark's the traveler.
17:03 hagarth jdarcy: ah ok
17:04 * jdarcy ponders how to apply Dynamo-style consistency to non-data operations.
17:04 hagarth johnmark: where art thou now?
17:06 johnmark hagarth: Paris :) for the OSAS all hands meeting
17:06 hagarth johnmark: awesome :)
17:06 johnmark heh
17:06 johnmark and then on to FOSDEM
17:06 johnmark it's my around-the-world tour
17:07 johnmark hagarth: literally, around the world
17:07 hagarth johnmark: yeah!
17:09 johnmark haha
17:17 spn joined #gluster-dev
18:03 inodb joined #gluster-dev
18:17 vpshastry left #gluster-dev
19:37 inodb joined #gluster-dev
19:39 kkeithley1 johnmark: catch if you can. See you tomorrow or Saturday. Where are you staying?
19:39 kkeithley1 s/catch/catch him/
20:03 kkeithley1 left #gluster-dev
21:05 inodb left #gluster-dev
23:25 hagarth joined #gluster-dev
23:47 jbrooks joined #gluster-dev

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