Perl 6 - the future is here, just unevenly distributed

IRC log for #gluster-dev, 2016-08-12

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

All times shown according to UTC.

Time Nick Message
00:04 lpabon joined #gluster-dev
00:31 purpleidea joined #gluster-dev
00:43 nigelb post-factum: no, I haven't. I need to submit a review request.
00:52 lpabon joined #gluster-dev
01:49 kdhananjay joined #gluster-dev
02:24 Manikandan joined #gluster-dev
02:56 spalai joined #gluster-dev
03:18 sanoj joined #gluster-dev
03:18 rafi joined #gluster-dev
03:30 shubhendu__ joined #gluster-dev
03:35 atinm joined #gluster-dev
03:39 penguinRaider joined #gluster-dev
03:42 magrawal joined #gluster-dev
03:48 itisravi joined #gluster-dev
03:51 hagarth kdhananjay: ping, there was a thought about recommending sharding instead of striping with 3.9
03:51 poornimag joined #gluster-dev
03:51 hagarth kdhananjay: how far are we from doing such a recommendation? i.e. for recommending sharding with general purpose workloads?
03:55 kdhananjay hagarth: so what features is sharding supposed to be compatible with for it to be "ready" for general purpose workloads?
03:56 kdhananjay hagarth: i did ask about this in the community long time ago to get some feedback, but the only response i got was from one person who said he would prefer that it works well with quota.
03:57 kdhananjay hagarth: if general purpose workloads still use single-writer to a given file, then it should mostly work fine without requiring any new changes to the code.
03:58 hagarth kdhananjay: multiple writers is what I am looking for
03:59 hagarth kdhananjay: already sent an email on -devel .. we can continue the conversation there too.
03:59 kdhananjay hagarth: ok. that would mostly require some form of locking to be introduced, especially to handle appending writes (O_APPEND), O_TRUNC etc.
03:59 kdhananjay hagarth: sure. i will take a look.
04:00 hagarth kdhananjay: quota would be nice to have but multiple writers is essential
04:00 kdhananjay hagarth: but it's not like striping handles all these problems in its current form, IIRC.
04:00 kdhananjay :)
04:00 hagarth kdhananjay: if striping were to be implemented the right way, that is how it would be doing :)
04:01 hagarth kdhananjay: striping is popular in HPC deployments where there are multiple writers & readers to the same file..
04:02 kdhananjay hagarth: yeah, so we may choose to keep striping still for 3.9, but it may have the same multiple writers problem as sharding.
04:03 hagarth since the parallel writes/reads can land on multiple chunks .. it can serve requests from various nodes .. thereby aggregating resources from multiple nodes and hence high performance
04:03 kdhananjay hagarth: im trying to get that common transaction infra in so that an inode write txn doesn't have to involve sharding first taking locks, and then afr again taking locks and finally doing the op and then unlocking twice.
04:04 kdhananjay because the number of network calls might increase if sharding also starts to take locks
04:04 kdhananjay hagarth: agree with you.
04:04 hagarth kdhananjay: right .. but when we want to suggest sharding as a replacement for striping, I would like it to handle all workloads without ifs and buts.. messaging that out from an user/admin perspective would be difficult
04:05 hagarth kdhananjay: that sounds like the right way to proceed (the common txn framework)
04:05 kdhananjay hagarth: yeah, ok i get that. the question is about how strict we need to be with synchronisation across those multiple writers if the writes may fall on the same set of shards concurrently
04:07 hagarth kdhananjay: offer the same behavior that we do in non-sharded cases.. we would need to serialize those operations
04:07 kdhananjay hagarth: ok.
04:13 shubhendu__ joined #gluster-dev
04:27 raghug joined #gluster-dev
04:29 sanoj joined #gluster-dev
04:49 nigelb Morning
05:01 kotreshhr joined #gluster-dev
05:03 ankitraj joined #gluster-dev
05:06 nbalacha joined #gluster-dev
05:07 nigelb nbalacha: Hey, can I mark tests/basic/quota.t as bad for netbsd on release-3.7
05:07 nigelb It's marked as bad on 3.8 and master.
05:08 nbalacha nigelb, I think it should be alright to do so
05:08 spalai left #gluster-dev
05:08 nbalacha Is it failing there?
05:08 nigelb lots of times :(
05:08 nigelb 16 failures in the last 2 weeks.
05:08 ppai joined #gluster-dev
05:09 nigelb post-factum and jiffin have been hitting it constantly.
05:10 karthik_ joined #gluster-dev
05:11 kotreshhr joined #gluster-dev
05:11 nigelb Aha, I can probably backport http://review.gluster.org/#/c/14136/
05:15 Manikandan joined #gluster-dev
05:17 jiffin joined #gluster-dev
05:18 nigelb jiffin: For your netbsd failures, I recommend backporting http://review.gluster.org/#/c/14136/ to 3.7
05:18 nigelb I'm trying to figure out how exactly to submit a request :)
05:18 ira_ joined #gluster-dev
05:20 ndarshan joined #gluster-dev
05:20 jiffin nigelb: k
05:20 jiffin thanks
05:20 jiffin i will backport that change
05:23 nishanth joined #gluster-dev
05:24 nigelb jiffin: I'm on it.
05:24 nigelb jiffin: just review it when I've posted it.
05:24 jiffin nigelb: http://review.gluster.org/#/c/15154/
05:24 nigelb ha
05:24 nigelb you're faster
05:25 jiffin i won the race :)
05:25 nigelb ha
05:25 jiffin please  review the change
05:25 nigelb <-- not a maintainer :)
05:25 glusterbot nigelb: <'s karma is now -13
05:26 nigelb ..
05:26 jiffin that's does not count, u can still review the change.
05:29 nigelb If nobody gives a +2 by the time regressions are posted
05:29 nigelb I'll do a +2 and merge it.
05:32 mchangir joined #gluster-dev
05:32 jiffin nigelb: okey
05:37 ramky joined #gluster-dev
05:39 hgowtham joined #gluster-dev
05:53 Bhaskarakiran joined #gluster-dev
05:57 kdhananjay joined #gluster-dev
06:06 skoduri joined #gluster-dev
06:06 Muthu joined #gluster-dev
06:09 pranithk1 joined #gluster-dev
06:10 aspandey joined #gluster-dev
06:10 spalai joined #gluster-dev
06:17 kshlm joined #gluster-dev
06:24 itisravi pranithk1: review requested for http://review.gluster.org/#/c/15146/ and http://review.gluster.org/#/c/15147/
06:25 msvbhat joined #gluster-dev
06:27 nigelb Huh. I may have caused a bug.
06:32 shubhendu joined #gluster-dev
06:34 * nigelb clears throat.
06:34 nigelb A netbsd-smoke job was stuck for a while.
06:35 nigelb I've cleared it out.
06:36 Manikandan joined #gluster-dev
06:43 ankitraj joined #gluster-dev
06:44 ankitraj joined #gluster-dev
06:51 atalur joined #gluster-dev
06:56 Bhaskarakiran_ joined #gluster-dev
06:58 Bhaskarakiran joined #gluster-dev
07:07 rastar joined #gluster-dev
07:07 rafi joined #gluster-dev
07:10 rafi1 joined #gluster-dev
07:19 aspandey joined #gluster-dev
07:25 ramky joined #gluster-dev
07:49 anoopcs joined #gluster-dev
07:51 karthik_ joined #gluster-dev
08:03 LinkRage joined #gluster-dev
08:05 LinkRage Is there a way to use GlusterFS without FUSE ?
08:14 pranithk1 LinkRage: you can use it with nfs/smb as well
08:14 pranithk1 rastar: skoduri: ^^
08:20 devyani7 joined #gluster-dev
08:21 * ndevos points to the chat in #gluster
08:21 LinkRage yep guys, ndevos answered all my questions already, thank you :)
08:29 ndevos kdhananjay: no more stripe, and sharding all the way? (please reply to vijays/mine email on gluster-devel)
08:30 kdhananjay ndevos: will do.
08:30 ndevos kdhananjay++ thanks - and let me know if you like to be added as a "deprecate stripe" feature owner ;-)
08:30 glusterbot ndevos: kdhananjay's karma is now 22
08:34 noobs joined #gluster-dev
08:39 sanoj joined #gluster-dev
08:51 nigelb Does anyone have time to review https://github.com/gluster/gluster​fs-patch-acceptance-tests/pull/53 and https://github.com/gluster/gluster​fs-patch-acceptance-tests/pull/54
09:02 ndevos sac`: have you ever used http://docs.ansible.com/ansib​le/gluster_volume_module.html ? any plans to improve it with whatever gdeploy has?
09:02 * ndevos thinks gdeploy is more advanced...
09:02 sac` ndevos, we looked at that initially.
09:02 sac` It had a lot of shortcomings.
09:03 sac` ndevos, so we decided to write our own. The reason for not building on is something I can't recollect.
09:04 misc nigelb: so I do not know jjb, so can't review that much, but that mean we get 1 build per job, but 1 jobs could still be on the same node ?
09:05 ndevos sac`: hmm, ok, I guess it would be good community behaviour if the gdeploy improvements/replacement gets submitted back to Ansible?
09:05 sac` ndevos, yes.
09:05 sac` It is on our todo list.
09:06 ndevos sac`: ah, nice to know, thanks!
09:06 sac` ndevos, we had a meeting with Ansible manager as well.
09:07 ndevos sac`: oh, good! I hope you'll be able to work close together with some of their devs
09:08 ndevos sac`: I hope we can push gdeploy a little more to our users, it would be great if you could send status/plan updates to gluster-devel every now and then
09:08 sac` ndevos, yes.
09:08 ndevos sac`: of course, the updates on what new features are included are already much appreciated!
09:09 sac` ndevos, thanks.
09:12 ndevos sac`: does gdeploy work fine with ansible-2.0? I could add the packages to the CentOS Storage SIG to get it some more testing
09:13 sac` Yes. I've ported it to Ansible 2.x
09:14 rastar joined #gluster-dev
09:14 ndevos cool!
09:18 nigelb misc: no, so I've specifically limited it so it's one job per node.
09:19 nigelb concurrently only one job will build on one node.
09:19 ira joined #gluster-dev
09:20 nigelb ndevos: sadly enough, getting the branch name is more challenging than I expected from Jenkins.
09:20 nigelb I might fetch it from Gerrit instead.
09:21 ndevos nigelb: hmm, isnt that in the environment variables for the run?
09:22 ndevos nigelb: that might mean fetching an other url, not only the /consoleAll one
09:22 poornimag joined #gluster-dev
09:25 nigelb ndevos: I'm using the Jenkins API.
09:25 nigelb I'll figure it out.
09:25 nigelb I want to minimize the API calls.
09:27 ndevos nigelb: I assume that the api provides a way to get the environment variables for a job too, should be pretty standard imho
09:28 nigelb ndevos: I can either get ALL THE INFORMATION or get a subset.
09:28 nigelb I'm getting a subset.
09:28 LinkRage left #gluster-dev
09:28 nigelb Getting the env variables down to a subset isn't bending to my will :)
09:28 ndevos heh, I wont help you with that :)
09:29 nigelb heh
09:29 nigelb I'll figure it out.
09:30 rraja joined #gluster-dev
09:34 nigelb misc: i've linked documentation for that particular invocation in the description of the pull request.
09:37 misc nigelb: yeah, I did read it :) but so it look ok, just pondering if I understand the doc correctly
09:39 nbalacha joined #gluster-dev
09:42 nigelb misc: The documentation is really excellent, to be honest. I've been loving how it's way easier to edit jobs en-masse now :)
09:44 misc nigelb: true, I can't really understand how people can use wbeinterface for  high scale work
09:47 ndevos kshlm: there is a bug in ./close-bugs.sh, it fails to close bugs (missing " after the bz-comment"
09:48 * ndevos sends a pull-request
09:51 ndevos see https://github.com/gluster/release-tools/pull/9
09:54 nigelb kshlm: Can you +2? http://review.gluster.org/#/c/15154/
09:54 kshlm ndevos++ Thanks
09:54 glusterbot kshlm: ndevos's karma is now 302
09:55 kshlm I didn't test the patch, and just assumed it would work. It was supposed to be a copy/paste job.
09:55 nigelb kshlm: with shell scripts, it's always better to test.
09:55 nigelb I can say so from experience :)
09:56 nigelb It's incredibly easy to make a mistake.
09:56 kshlm nigelb, You want me to submit as well?
09:56 nigelb yes, please
09:57 kshlm (Just want to note that '/tests/' is the domain of rastar and msvbhat)
09:57 ndevos hey, nigelb did you look at that pull request I just posted? Untested shell scripts :-/
09:57 nigelb ndevos: I was looking at it when kshlm merged it :)
09:57 ndevos nigelb: now guess who broke it ;-)
09:57 nigelb haha
09:58 nigelb post-factum / jiffin - a rebase and netbsd recheck should get that review a green netbsd (hopefully)
10:00 nigelb one of the reasons I prefer python to shell script is because I can write unit tests for business logic.
10:00 post-factum nigelb: thanks, will try now
10:01 jiffin nigelb:once again thank you
10:01 nigelb I didn't do much!
10:07 hgowtham joined #gluster-dev
10:12 nigelb misc: have you got that repository request? Or shall I grab it?
10:13 misc nigelb: better grab it, I am still in a bus to the airport for now
10:13 misc I can fix urgent stuff, but connectivity is crappy
10:13 nigelb no worries, I got it.
10:22 spalai left #gluster-dev
10:24 justinclift joined #gluster-dev
10:26 justinclift kkeithley: Yep?  It's sitting next to me at the moment, installing MSVC2013 into a VM, for building stuff on. ;)
10:28 justinclift Question to all: With GlusterFS being dual licensed, if we have to "pick one" for this, which one?  https://github.com/Kickball/awesome-sel​fhosted/pull/647#issuecomment-239368503
10:43 msvbhat joined #gluster-dev
11:01 rastar joined #gluster-dev
11:03 poornimag joined #gluster-dev
11:25 spalai joined #gluster-dev
11:39 shubhendu_ joined #gluster-dev
11:44 pranithk1 joined #gluster-dev
12:01 atalur joined #gluster-dev
12:04 shubhendu__ joined #gluster-dev
12:18 post-factum nigelb++ jiffin++ for disabling quota test on netbsd. works now
12:18 glusterbot post-factum: nigelb's karma is now 24
12:18 glusterbot post-factum: jiffin's karma is now 49
12:19 jiffin post-factum: :)
12:27 mchangir joined #gluster-dev
12:32 shubhendu_ joined #gluster-dev
12:37 atinm joined #gluster-dev
12:39 pranithk1 joined #gluster-dev
12:39 pranithk1 joined #gluster-dev
12:48 mchangir joined #gluster-dev
12:57 shubhendu__ joined #gluster-dev
12:58 rastar joined #gluster-dev
12:59 nigelb post-factum: woo!
13:12 shubhendu_ joined #gluster-dev
13:12 dlambrig_ joined #gluster-dev
13:18 shubhendu_ joined #gluster-dev
13:30 shubhendu__ joined #gluster-dev
13:43 rastar joined #gluster-dev
13:50 msvbhat joined #gluster-dev
13:52 shubhendu__ joined #gluster-dev
13:58 ramky joined #gluster-dev
14:00 shyam joined #gluster-dev
14:06 lpabon joined #gluster-dev
14:18 ramky joined #gluster-dev
14:26 post-factum ndevos: http://review.gluster.org/#/c/13488/ this one waits for your review
14:29 Muthu joined #gluster-dev
14:31 ndevos post-factum: oh, I'm sure it does, the number of patches that I need to review is longer than I can see on my monitor :-
14:31 ndevos \
14:33 post-factum ndevos: mail your manager to buy new and shiny vertical monitor for you :)
14:34 post-factum ndevos: hint — you may stack 2 vertical monitors to make it even longer
14:35 ndevos 2 vertical monitors on top? I have 2x 22" monitors now, that feels huge already
14:35 ndevos (one is vertical, for emails)
14:36 post-factum ndevos: also, you may decrease font size
14:37 ndevos :P
14:39 hagarth ndevos: what is your font size now?
14:40 post-factum ndevos: https://twitter.com/nixpani​c/status/577378599071805440 this?
14:40 post-factum ndevos: nice setup
14:41 hagarth post-factum: the font size looks huuuuge
14:41 hagarth ndevos should indeed cut down on it to review faster
14:41 post-factum ye, ndevos, you could definitely fit more
14:43 ndevos post-factum: hah, yes
14:57 raghug joined #gluster-dev
15:02 wushudoin joined #gluster-dev
15:17 mchangir joined #gluster-dev
15:23 ramky joined #gluster-dev
15:38 mchangir joined #gluster-dev
15:45 shyam joined #gluster-dev
16:00 pranithk1 joined #gluster-dev
16:03 shyam joined #gluster-dev
16:03 justinclift hagarth: With GlusterFS being dual licensed, if we have to "pick one" for this, which one?   https://github.com/Kickball/awesome-sel​fhosted/pull/647#issuecomment-239368503
16:04 * justinclift is thinking it might be better to just remove both the license terms.  Not every project on there lists them.
16:08 glustin joined #gluster-dev
16:27 shyam joined #gluster-dev
16:34 mchangir nigelb, are you working?
16:42 dlambrig left #gluster-dev
16:52 pranithk1 justinclift: hey! how are you doing?
16:52 pranithk1 justinclift: long time!!
16:58 justinclift pranithk1: I'm good.  Building packages for a new DB4S release at the moment.
16:58 justinclift (our first beta in ages)
16:58 justinclift pranithk1: How's stuff in Gluster land? :D
16:58 pranithk1 justinclift: small file perf increasing slow and steady :-)
16:59 justinclift Hopefully not too slow ;)
17:00 pranithk1 justinclift: well, write went from 5->3 fops because of compound fops. We are developing -ve lookup preventing xlator for new file creations which we need to test how much it will improve
17:02 justinclift Compounds ops sounds like a good idea.  Transaction like?
17:02 justinclift eg either fails or succeeds, no "1 part succeeded, then the next failed" ?
17:03 nigelb mchangir: nope :)
17:17 pranithk1 justinclift: hmm.. not really, you can send two fops to a brick with one network hop...
17:17 pranithk1 justinclift: I mean one network fop...
17:19 hagarth joined #gluster-dev
17:20 pranithk1 justinclift: gotta go now, it is nice seeing you after a long time... bbye. All the best with your release!
17:20 justinclift tx :)
17:21 hagarth justinclift: will comment on that one soon
17:34 justinclift hagarth: Thx
17:34 mchangir left #gluster-dev
17:36 mchangir_ joined #gluster-dev
17:49 circ-user-ZfidW joined #gluster-dev
18:43 kotreshhr joined #gluster-dev
18:50 glustin joined #gluster-dev
19:46 PotatoGim joined #gluster-dev
19:59 shyam joined #gluster-dev
20:18 glustin joined #gluster-dev
21:09 lpabon joined #gluster-dev
21:14 lpabon joined #gluster-dev
21:40 lpabon joined #gluster-dev
22:27 lpabon joined #gluster-dev
22:58 lpabon joined #gluster-dev
23:12 hagarth joined #gluster-dev
23:44 lpabon joined #gluster-dev

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