Perl 6 - the future is here, just unevenly distributed

IRC log for #gluster-dev, 2015-05-22

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

All times shown according to UTC.

Time Nick Message
00:02 ppai joined #gluster-dev
00:32 dlambrig joined #gluster-dev
03:02 sakshi joined #gluster-dev
03:16 rjoseph joined #gluster-dev
03:33 hagarth joined #gluster-dev
03:38 pranithk joined #gluster-dev
03:41 kanagaraj joined #gluster-dev
03:54 atinmu joined #gluster-dev
03:58 dlambrig joined #gluster-dev
04:00 dlambrig left #gluster-dev
04:10 kdhananjay joined #gluster-dev
04:17 jiffin joined #gluster-dev
04:17 shubhendu joined #gluster-dev
04:31 pranithk xavih: ndevos: I initiated a thread on mem_0filled/iov_0filled.
04:32 spandit joined #gluster-dev
04:33 rafi joined #gluster-dev
04:35 schandra joined #gluster-dev
04:41 deepakcs joined #gluster-dev
04:55 pppp joined #gluster-dev
04:58 shubhendu joined #gluster-dev
04:58 ndarshan joined #gluster-dev
05:07 gem joined #gluster-dev
05:11 hgowtham joined #gluster-dev
05:26 atalur joined #gluster-dev
05:26 tigert morning
05:28 tigert spot: maybe we could have a github repo for presentations? :P not ideal tool, but works
05:29 tigert https://github.com/puppetlabs/showoff < oh nice
05:29 tigert Showoff allows you to author your presentation slides in Markdown, then organize them with a showoff.json file.
05:29 tigert markdown is the docker of text content!
05:29 dlambrig joined #gluster-dev
05:38 nkhare joined #gluster-dev
05:39 ashiq joined #gluster-dev
05:42 ashish joined #gluster-dev
05:44 hagarth joined #gluster-dev
05:45 rafi1 joined #gluster-dev
05:54 overclk joined #gluster-dev
06:01 aravindavk joined #gluster-dev
06:15 rafi joined #gluster-dev
06:21 poornimag joined #gluster-dev
06:21 Gaurav_ joined #gluster-dev
06:42 pranithk joined #gluster-dev
06:43 overclk hagarth (and everyone else): need reviews on 10763 and 10832 please.
06:44 hagarth overclk: yes, on my list. looking more into regression tests atm.
06:44 raghu joined #gluster-dev
06:46 Guest15424 joined #gluster-dev
06:47 Anjana joined #gluster-dev
06:48 aravindavk joined #gluster-dev
06:52 spalai joined #gluster-dev
06:54 spalai1 joined #gluster-dev
06:59 overclk thanks hagarth.
07:04 spandit joined #gluster-dev
07:05 kshlm joined #gluster-dev
07:06 Guest15424 joined #gluster-dev
07:23 pranithk xavih: All the tests that expect for 0 xattrs to be present on the brick are failing :-(
07:24 xavih pranithk: I thought about that and was thinking if adding size==-1 as a condition to update the xattr
07:25 xavih pranithk: this would change ctime, and the effect from the client side would be the same. I didn't thought with tests cases that directly access the bricks :(
07:25 pranithk xavih: i was thinking the same thing :-). But it is going to create the same time problem them
07:25 pranithk xavih: yeah
07:25 pranithk xavih: correct thing is to change the tests
07:26 pranithk xavih: There are too many.
07:27 xavih pranithk: there's another possibility, but it's a hack and probably not much clean...
07:27 kdhananjay joined #gluster-dev
07:28 xavih pranithk: we could do special handling when size==-1 and xattrop value is 0: create the xattr and restore previous ctime
07:28 pranithk xavih: no, lets fix the tests
07:29 xavih pranithk: ok. I'll upload a new patchset for the cooperative locking (already done, only need to upload) and I'll start with this
07:29 pranithk xavih: wait
07:29 pranithk xavih: I changed run-tests.sh to remove stop at the first failure.
07:30 pranithk xavih: so it will run the full test suit. We can see what all tests have this problem. Fix all of them and resend
07:30 xavih pranithk: ok
07:34 xavih pranithk: I've updated the cooperative locking patch because the assert in ec_lock_unfreeze() was really necessary. I've even added a check for lock->owner == NULL. This is needed to avoid two fops progressing at the same time
07:34 xavih pranithk: This could cause list (memory) corruption and failures on another assert
07:35 xavih pranithk: I've made some changes to guarantee that these conditions are met when ec_lock_unfreeze() is called
07:35 pranithk xavih: Cool, I will re-test these things.
07:38 pranithk xavih: about you mail, using ia_type and lock_type to choose 'version' properly is necessary for getting the version correctly right?
07:39 xavih pranithk: why ? setattr should always use metadata version independently if it's a directory or a regular file, right ?
07:40 hchiramm_ Cron <root@supercolony> /bin/cp -R -f /var/www/middleman_website/master/* /var/www/staging/ tigert
07:40 hchiramm_ I am keep getting above notification..
07:41 hchiramm_ and quite annoying.. :(
07:41 pranithk xavih: xattrop is going to happen after taking inodelks, but based on status of previous entry operation, version updating of 'data' part could be in progress leading to mismatching xdata....
07:41 pranithk xavih: because entry operations happen under entry locks
07:43 xavih pranithk: oh, I see. I'll incorporate this logic...
07:45 xavih pranithk: this only happens for (f)xattrop when used as a lookup for size and version values, right ?
07:46 pranithk xavih: that is where we observed it for now...
07:46 pranithk xavih: But should happen even when it does update_size_version?
07:47 xavih pranithk: I think that's not a problem because xattrop are atomic, and the returned value is not checked after updating
07:47 pranithk xavih: xattrop also does combine of xdata...?
07:47 pranithk xavih: which is failing...
07:48 pranithk xavih: because the xattrs don't match because of parallel entry/metadata operations on the same file.
07:48 xavih pranithk: but this will be the same case as the first one
07:48 pranithk xavih: same directory not file :-)
07:48 pranithk xavih: You mean get_size_version?
07:49 pranithk xavih: Yes if we fix get_size_version this will also be fixed. Because you if I understand correctly, were fixing it in xattrop merging the values...
07:49 tigert hchiramm_: yeah we are looking into it
07:49 xavih pranithk: I was thinking to do the changes in ec_combine_xattrop. This way it will only consider the correct value
07:49 xavih pranithk: yes
07:49 pranithk xavih: Yeah :-)
07:50 pranithk xavih: But you will need to change the signature of combine now just for this :-/
07:50 pranithk xavih: Oh not really
07:51 pranithk xavih: It already takes fop as parameter..
07:51 xavih pranithk: yes. I think I've all the necessary things there :)
07:51 pranithk xavih: okay sir! we are good to go.
07:56 pranithk xavih: Sent you a mail, about my status...
08:01 tigert hchiramm_: mail nags should be gone now
08:04 hchiramm_ tigert++ thanks
08:04 glusterbot hchiramm_: tigert's karma is now 10
08:23 xavih pranithk: to change the behavior of xattrop, I need to have each xattrop call associated with a specific lock. I don't have this information in the cooperative locking patch. I added this info to the patch http://review.gluster.org/10852/. It's ok if I modify this patch instead of the cooperative locking ?
08:24 anrao joined #gluster-dev
08:25 hagarth pranithk: will you look into the newly reported failure on ec-5-1.t ?
08:26 hagarth jiffin: can you please help with mount-nfs-auth.t failure?
08:26 jiffin sorry hagarth , i didn't notice that issue
08:27 jiffin hagarth: where can i find the details about the issue?
08:27 hagarth jiffin: thanks, if you are looking into .. please update the spurious failures etherpad.
08:27 hagarth jiffin: you can find details in shyam's email on gluster-devel
08:28 jiffin hagarth: k
08:28 pranithk hagarth: It is based on old run before merge of the EIO fix I sent
08:28 pranithk hagarth: The run happened on 20th
08:28 pranithk patch merged on 20th as well
08:28 hgowtham joined #gluster-dev
08:29 soumya joined #gluster-dev
08:29 tigert hagarth: what is blog.gluster.org btw?
08:29 tigert hagarth: is it a wiki or what? /me has no idea what that is and if it is used for anything
08:29 hagarth pranithk: fantastic!
08:30 pranithk xavih: totally :-)
08:30 hagarth pranithk: maybe update the thread and again nuke the test from is_bad_test() in one of your subsequent patches.
08:30 hagarth tigert: I think blog.gluster.org should point to planet gluster now
08:30 pranithk hagarth: But we told that xavih http://review.gluster.org/10852 also is a candidate for fixing the EIO problems. So may be as part of that we can nuke it.
08:31 pranithk xavih: I will be away for an hour now...
08:31 tigert hagarth: is blog.gluster.org the wordpress?
08:32 tigert or at least there is a wordpress in blog.gluster.org/blog/
08:32 tigert that should die, right?
08:32 tigert but is that still wired to the social media stuff?
08:33 hagarth tigert: that could .. but do we need an official blog for gluster?
08:33 tigert we kind of do have one
08:33 tigert http://www.gluster.org/news/
08:33 tigert that *is* a blog
08:33 tigert you can write articles
08:33 tigert and we can add the feed to planet
08:34 tigert http://www.gluster.org/news/feed.xml <- the feed
08:34 tigert that would be the official blog thing
08:34 tigert hagarth: do you have the glusterweb repo checked out in a machine with ruby?
08:34 hagarth tigert: cool
08:35 hagarth tigert: yes, my laptop
08:35 tigert cd g,h
08:35 tigert ermh
08:35 tigert cd glusterweb,
08:35 tigert ./setup.sh (if you havent before)
08:35 tigert ./run-server.sh
08:36 tigert browse to http://0.0.0.0:4567/admin/
08:37 tigert you still need to push the files to git, but it helps you with writing blog posts
08:38 hagarth tigert: cool
08:41 tigert https://dl.dropboxusercontent.com/u/​2730444/foss/gluster-post-admin.png
08:43 tigert it is something that Garrett did for manageiq site, and since we use the same framework we can reuse bits
08:43 tigert you can alternatively write direct markdown files too
08:43 kdhananjay joined #gluster-dev
08:45 hagarth tigert: nice
08:52 gem joined #gluster-dev
08:53 nishanth joined #gluster-dev
09:04 ashiq joined #gluster-dev
09:06 aravindavk joined #gluster-dev
09:13 anekkunt joined #gluster-dev
09:14 gem joined #gluster-dev
09:17 atalur joined #gluster-dev
09:19 kdhananjay joined #gluster-dev
09:25 pranithk joined #gluster-dev
09:26 ashiq joined #gluster-dev
09:34 ira joined #gluster-dev
09:37 pranithk xavih: for (f)xattrop patch Except that test we saw the failure, nothing else failed in the regression run. I went ahead and posted the fix I sent you in mail...
09:38 xavih pranithk: ok. Great :)
09:38 xavih pranithk: Thanks
09:45 jiffin1 joined #gluster-dev
09:52 atalur joined #gluster-dev
09:54 rraja joined #gluster-dev
09:59 atinmu joined #gluster-dev
10:21 rraja joined #gluster-dev
10:30 pranithk hagarth: xavih: ec test cases started failing again, this time I think it is because of self-heal. I need to leave now to catch a train. Will take a look at them with priority tomorrow morning. Cya
10:30 jiffin joined #gluster-dev
10:42 atalur joined #gluster-dev
10:47 jiffin joined #gluster-dev
10:57 atinmu joined #gluster-dev
11:04 aravindavk joined #gluster-dev
11:07 jiffin1 joined #gluster-dev
11:14 rraja joined #gluster-dev
11:54 atalur joined #gluster-dev
11:57 gem joined #gluster-dev
11:57 nishanth joined #gluster-dev
12:53 ira joined #gluster-dev
12:59 hagarth joined #gluster-dev
13:00 jiffin joined #gluster-dev
13:05 jiffin1 joined #gluster-dev
13:14 rraja joined #gluster-dev
13:16 nkhare joined #gluster-dev
13:23 shyam joined #gluster-dev
13:47 ndevos raghu: a netbsd/freebsd smoke test for release-3.6 failed, do you know about this? http://build.gluster.org/job​/netbsd6-smoke/6402/console
13:59 firemanxbr joined #gluster-dev
14:09 wushudoin joined #gluster-dev
14:09 wushudoin joined #gluster-dev
14:14 vimal joined #gluster-dev
14:30 kdhananjay joined #gluster-dev
14:31 rafi ndevos: Hi
14:32 ndevos hello rafi!
14:33 rafi ndevos: how are you ?
14:33 ndevos rafi: I'm good, how are you?
14:33 rafi ndevos: I'm doing good
14:33 rafi ndevos: I had doubt regarding the patch http://review.gluster.org/#/c/107​99/1/xlators/cluster/dht/src/tier.c
14:34 ndevos rafi: yes, that is rather ugly :)
14:34 ndevos rafi: what is your doubt?
14:35 rafi ndevos: what happens if the version is changed ?
14:35 rafi ndevos: All other dynamic library are loading using the name.so format
14:36 rafi ndevos: I guess name.so is actually a symbolic link to the actual version
14:36 rafi ndevos: I'm not sure
14:37 rafi ndevos: what do you think ?
14:37 ndevos rafi: yes, the .so is a symlink for the libgfdb.so.0 version, but only if you have glusterfs-devel installed (or build from sources)
14:37 ndevos rafi: libgfdb.so does not have a fixed API yet, so it would need some library versioning to prevent issues
14:38 ndevos rafi: I would rather see the version set in configure.ac, and define it in config.h with AC_DEFINE
14:39 * rafi is checking
14:40 rafi ndevos: I don't have much knowledge about this, I'm sorry if it was a stupid doubt  :)
14:40 ndevos rafi: no problem :)
14:40 ndevos rafi: joseph asked about it a few days ago :D
14:41 rafi ndevos: Ok
14:42 rafi ndevos: thanks for clarifying the doubt
14:42 rafi ndevos++
14:42 glusterbot rafi: ndevos's karma is now 130
14:43 ndevos rafi: thing is, the symlink in needed for developers, they link their binaries with: gcc -lgfdb ....
14:43 shaunm_ joined #gluster-dev
14:43 ndevos and that -lgfdb resolves libgfdb.so, and links with the actual version of libgfdb
14:44 dlambrig joined #gluster-dev
14:44 rafi ndevos: ok
14:45 rafi ndevos: I should study more about this :)
14:46 ndevos rafi: it helps to know how things work :)
14:47 rafi ndevos: definitely
14:48 ndevos rafi: ah, lol! I just notice I sent that patch :D
14:48 * ndevos wanted to comment on it, but will not -1 it now
14:49 rafi ndevos: :) let it go
14:51 ndevos rafi: +1 it!
14:52 atinmu joined #gluster-dev
14:52 rafi ndevos: i put a +1 on it
14:53 ndevos rafi++ thank you :)
14:53 glusterbot ndevos: rafi's karma is now 13
14:53 rafi ndevos: :)
15:13 atalur joined #gluster-dev
15:33 dlambrig joined #gluster-dev
15:42 atinmu joined #gluster-dev
15:57 atalur joined #gluster-dev
16:29 atalur_ joined #gluster-dev
16:42 dlambrig joined #gluster-dev
17:14 hagarth xavih: pranith has observed some corruption in ec_lock() during his train journey :)
17:16 hagarth xavih: he also mentioned that a hang is observed when parallel writes happen
17:27 xavih hagarth: I'm working on this. I've identified a possible cause. I need to do some changes...
17:28 hagarth xavih: ok, thanks.
17:44 ppai joined #gluster-dev
17:52 hagarth joined #gluster-dev
17:52 rafi joined #gluster-dev
18:07 _Bryan_ joined #gluster-dev
18:10 ppai joined #gluster-dev
18:10 ppai joined #gluster-dev
18:15 Gaurav_ joined #gluster-dev
18:45 shaunm_ joined #gluster-dev
19:53 ppai joined #gluster-dev
19:55 dlambrig joined #gluster-dev
20:38 dlambrig fatal: remote error: access denied or repository not exported: /glusterfs
20:38 dlambrig anyone know what that means , when try to commit a patch using rfc.sh
20:39 dlambrig left #gluster-dev
21:53 shaunm_ joined #gluster-dev
23:09 dlambrig joined #gluster-dev
23:18 dlambrig left #gluster-dev

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