Perl 6 - the future is here, just unevenly distributed

IRC log for #gluster-dev, 2015-07-01

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

All times shown according to UTC.

Time Nick Message
00:29 [o__o] joined #gluster-dev
01:21 aravindavk joined #gluster-dev
02:03 dlambrig_ left #gluster-dev
02:12 ndk joined #gluster-dev
02:22 pppp joined #gluster-dev
02:28 soumya_ joined #gluster-dev
03:00 kdhananjay joined #gluster-dev
03:27 overclk joined #gluster-dev
03:35 sakshi joined #gluster-dev
03:38 atinm joined #gluster-dev
03:54 kanagaraj joined #gluster-dev
03:57 vmallika joined #gluster-dev
03:59 itisravi joined #gluster-dev
03:59 nishanth joined #gluster-dev
04:22 gem joined #gluster-dev
04:28 suliba joined #gluster-dev
04:33 deepakcs joined #gluster-dev
04:34 nbalacha joined #gluster-dev
04:42 shubhendu joined #gluster-dev
04:44 gem joined #gluster-dev
04:50 vimal joined #gluster-dev
04:50 jiffin joined #gluster-dev
04:58 ppai joined #gluster-dev
05:00 krishnan_p joined #gluster-dev
05:02 ashish joined #gluster-dev
05:05 pppp joined #gluster-dev
05:06 sakshi joined #gluster-dev
05:10 hgowtham joined #gluster-dev
05:10 josferna joined #gluster-dev
05:14 ashiq joined #gluster-dev
05:16 Manikandan_ joined #gluster-dev
05:17 kshlm joined #gluster-dev
05:18 pranithk joined #gluster-dev
05:18 ndarshan joined #gluster-dev
05:23 spandit joined #gluster-dev
05:26 rjoseph joined #gluster-dev
05:30 rafi joined #gluster-dev
05:46 Bhaskarakiran joined #gluster-dev
05:47 hagarth joined #gluster-dev
05:49 vmallika joined #gluster-dev
05:50 anekkunt joined #gluster-dev
05:52 anekkunt krishnan_p, please can you this patch  http://review.gluster.org/#/c/11120/ , it has all acks
05:53 anekkunt *merge
05:53 krishnan_p anekkunt, I will take a look later today.
05:54 anekkunt krishnan_p, ok
05:55 kdhananjay joined #gluster-dev
06:00 kdhananjay joined #gluster-dev
06:00 atalur joined #gluster-dev
06:02 nbalacha joined #gluster-dev
06:02 atalur joined #gluster-dev
06:07 raghu joined #gluster-dev
06:12 Gaurav__ joined #gluster-dev
06:31 saurabh joined #gluster-dev
06:40 aravindavk joined #gluster-dev
06:45 gem joined #gluster-dev
06:54 gem_ joined #gluster-dev
07:04 spalai joined #gluster-dev
07:08 rjoseph joined #gluster-dev
07:16 vmallika joined #gluster-dev
07:19 atinm joined #gluster-dev
07:19 pppp joined #gluster-dev
07:22 pranithk joined #gluster-dev
07:28 anrao joined #gluster-dev
07:37 atalur joined #gluster-dev
07:57 pranithk joined #gluster-dev
08:04 soumya joined #gluster-dev
08:16 atinm joined #gluster-dev
08:20 atalur joined #gluster-dev
08:22 pranithk xavih: Just added you as reviewer for all the self-heal throttling patches.
08:23 pranithk xavih: I also need to talk about 'ctx->bad' in inode-ctx. Ping me when you get some time.
08:32 soumya joined #gluster-dev
08:55 xavih pranithk: hi. I'll review the patches this morning
08:55 pranithk xavih: Found anything about the corruption on appends?
08:56 pranithk xavih: Bhaskarakiran is testing the build with these patches to see if it fixes the memory problems, i.e. 1) OOM killers because of too many heals 2) Will calloc/free still hang after this..
08:57 xavih pranithk: not yet. Bhaskarakiran has sent me more information to be able to analyze exactly what is happening. Unfortunately I've been unable to reproduce it, so I can only work based on the case Bhaskarakiran found
08:57 pranithk xavih: I have a feeling that alloca giving undefined behavior at memory limits could be the cause for calloc/free hang.... but I am not sure
08:58 pranithk xavih: He is seeing it consistently. If you want to add some extra information as a patch, we can get more information to you by providing a build with that info to Bhaskar. Let me know
08:58 xavih pranithk: could be. I personally don't like a function that can fail without notice
08:59 ppai joined #gluster-dev
09:00 xavih pranithk: I know he can reproduce it quite easily. It's really weird that I'm unable to get a single failure. There must be some other factor that we are not seeing...
09:01 ashiq thanks overclk++
09:01 glusterbot ashiq: overclk's karma is now 14
09:04 rgustafs joined #gluster-dev
09:08 pppp joined #gluster-dev
09:18 shubhendu joined #gluster-dev
09:18 nishanth joined #gluster-dev
09:19 aravindavk joined #gluster-dev
09:19 ndarshan joined #gluster-dev
09:26 krishnan_p ndevos, would http://review.gluster.org/#/c/11352/ be of interest for release-3.7? This patch has been merged upstream. I was thinking of abandoning this change for now and cherry-pick it on a need basis?
09:42 krishnan_p pranithk, ^^, thoughts?
09:43 krishnan_p pranithk, remember our discussion on how this change is intrusive and needs soak time on master before we cherry-pick this onto a stable branch?
09:43 krishnan_p pranithk, so, I am inclined to abandon it, if are not taking it any time soon, for the above reason.
09:46 pranithk krishnan_p: I was waiting for NetBSD run, we can merge it post that...
09:49 krishnan_p pranithk, ah OK.
09:50 krishnan_p pranithk, I will retrigger the NetBSD run then
09:56 pranithk krishnan_p: cool, sir!
10:08 surabhi_ joined #gluster-dev
10:08 rjoseph joined #gluster-dev
10:18 ndarshan joined #gluster-dev
10:19 raghu itisravi: is there any bug for the bad file heal issue??
10:20 nishanth joined #gluster-dev
10:20 itisravi raghu: what is bad file heal issue?
10:21 soumya joined #gluster-dev
10:21 shubhendu joined #gluster-dev
10:21 raghu itisravi: the file marked as bad by scrubber being removed from the backend and self-head daemon not being able to heal that object from the other replica due to bit-rot-stub blocking read/write operations on that inode
10:22 itisravi raghu: Ah sorry got it. I don't think Rajesh Reddy has raised a bug yet.
10:23 raghu itisravi: ok.
10:28 an joined #gluster-dev
10:48 itisravi raghu: pm
10:54 ppai joined #gluster-dev
11:12 ndevos krishnan_p, pranith: I do not see a problem in merging http://review.gluster.org/11352 in 3.7, if we notice a breakage, we'll fix it
11:13 ndevos ah! overclk, krishnan_p, pranithk, itisravi, raghu: can one of you host the weekly meeting?
11:13 ndevos or anyone else of course :)
11:15 soumya joined #gluster-dev
11:19 ndevos atinm: ^
11:25 kdhananjay1 joined #gluster-dev
11:27 pousley joined #gluster-dev
11:29 ira joined #gluster-dev
11:32 rafi1 joined #gluster-dev
11:35 kdhananjay joined #gluster-dev
11:38 ndevos overclk, krishnan_p, pranithk, itisravi, raghu, atinm: can one of you host the weekly meeting?
11:38 ndevos we really need to put something like a schedule/roster in place for this....
11:42 itisravi ndevos: sorry...I will have to give it a pass :(
11:43 pousley_ joined #gluster-dev
11:44 ndevos itisravi: can you physically poke someone over there?
11:44 itisravi ndevos: Sure, will do that :)
11:45 overclk itisravi, ndevos me too, pass today.
11:47 dlambrig_ joined #gluster-dev
11:48 xavih pranithk: ping, I'm reviewing the patch to limit background self-heals
11:50 * itisravi can't seem to find any volunteers.
11:57 ndevos itisravi: hmm, nobody?
11:58 ndevos kkeithley: do you want to host the meeting in 2 minutes? :)
11:58 itisravi ndevos: no one who wants to host..
11:58 * ndevos sadly shakes his head
11:59 kkeithley is it in two minutes or an hour?
12:00 ndevos kkeithley: more or less *now*
12:00 kkeithley hagarth sent an update invite that changed the time.
12:00 kkeithley (but the year was wrong and it erased all my calendar entries)
12:01 kkeithley changed the time to 9:00 EDT (13:00 UTC)
12:01 ndevos kkeithley: yeah, hagarth had an AI to delete the invite
12:02 * ndevos feels he's running around like a headless chicken with all the different bugs people think need fixing *NOW* *NOW* *NOW*
12:04 kkeithley meeting in #gluster-meeting
12:04 * kkeithley tries to remember meeting commands
12:04 kkeithley Gluster Community Meeting _now_ in #gluster-meeting
12:06 rafi joined #gluster-dev
12:06 anmolb joined #gluster-dev
12:09 jrm16020 joined #gluster-dev
12:20 an joined #gluster-dev
12:34 mribeirodantas joined #gluster-dev
13:16 ndevos kkeithley++ thanks for hosting the meeting!
13:16 glusterbot ndevos: kkeithley's karma is now 78
13:22 kbyrne joined #gluster-dev
13:23 kdhananjay joined #gluster-dev
13:25 pranithk xavih: hey!
13:25 pranithk xavih: Found any bugs? I don't see any updates on the patch...
13:26 surabhi_ joined #gluster-dev
13:26 xavih pranithk: I just have one comment. It's about making the maximum number of background self-heals and the number of queued heals configurable
13:27 xavih pranithk: I think it should be configurable, but I'm ok with this patch if it's added in another patch
13:29 sakshi joined #gluster-dev
13:29 sberry joined #gluster-dev
13:40 ndevos raghu, pranithk: http://review.gluster.org/11144 contains some posix (+acl) aspects, could you have a look at that soon?
13:59 wushudoin joined #gluster-dev
13:59 pranithk xavih: what should be the limits?
14:00 pranithk xavih: each syncop heal takes 2MB stack
14:00 xavih pranithk: I'm ok with the limits, but I think it should be configurable
14:01 xavih pranithk: the question is if this change is made in this patch or on another one
14:02 pranithk xavih: Hmm... another one :-)
14:03 xavih pranithk: ok
14:03 dlambrig_ joined #gluster-dev
14:04 pranithk xavih: Remember the calloc hang that we were observing? if we disable background heals it is not happening at all. That means I introduced some bad bug when I re-wrote self-heal...
14:05 shyam joined #gluster-dev
14:09 pranithk xavih: http://review.gluster.org/11472 is fine as well?
14:13 xavih pranithk: all patches reviewed
14:13 pranithk xavih: Today we made some progress.... I am going to look at healing code carefully to see if I am screwing up something... Will update if I find something
14:14 pousley joined #gluster-dev
14:14 pranithk xavih: Xavi the return is only happening when there is no frame...
14:15 xavih pranithk: even in that case, an EINVAL should be passed to the callback
14:15 xavih pranithk: or some other error if you prefer
14:16 xavih pranithk: oops... I think I read it completely wrong...
14:16 xavih pranithk: let me think again...
14:17 pranithk xavih: nothing should be waiting there, because there is no frame...
14:18 xavih pranithk: there could be a callback even if frame is NULL
14:18 pranithk xavih: Ah! you are right
14:18 pranithk xavih: I can change that :-)
14:22 xavih pranithk: I think the comment is ok. The callback should be called
14:24 shyam joined #gluster-dev
14:26 pranithk xavih: Yes, it is. okay I am leaving for home now. see you tomorrow.
14:33 kanagaraj joined #gluster-dev
14:54 ilbot3 joined #gluster-dev
14:54 Topic for #gluster-dev is now Gluster Development Channel - http://gluster.org | For general chat go to #gluster | Patches - http://review.gluster.org/ | Channel Logs - https://botbot.me/freenode/gluster-dev/ & http://irclog.perlgeek.de/gluster-dev/
15:04 josferna joined #gluster-dev
15:19 tdasilva joined #gluster-dev
15:20 aravindavk joined #gluster-dev
15:22 soumya joined #gluster-dev
15:24 kkeithley ,,(ppa)
15:31 spalai left #gluster-dev
15:49 an joined #gluster-dev
15:58 deepakcs joined #gluster-dev
16:05 jobewan joined #gluster-dev
16:41 dlambrig_ joined #gluster-dev
16:49 vmallika joined #gluster-dev
16:53 jiffin joined #gluster-dev
16:56 pppp joined #gluster-dev
17:11 jiffin joined #gluster-dev
17:14 vmallika joined #gluster-dev
17:18 soumya joined #gluster-dev
17:21 vmallika joined #gluster-dev
17:29 vmallika1 joined #gluster-dev
17:30 firemanxbr joined #gluster-dev
17:32 hagarth joined #gluster-dev
17:35 jiffin joined #gluster-dev
17:36 vimal joined #gluster-dev
17:51 jiffin joined #gluster-dev
17:59 pppp joined #gluster-dev
18:08 rafi joined #gluster-dev
18:10 rafi joined #gluster-dev
18:15 ndevos xavih++ thanks for the ec/seek() patch! I'll have a look at it later this week
18:15 glusterbot ndevos: xavih's karma is now 18
18:27 jiffin ping ndevos  i have doubt regarding http://review.gluster.org/#/c/11144/
18:27 ndevos hi jiffin
18:27 jiffin in one of ur comments
18:28 jiffin ndevos hi
18:28 ndevos jiffin: you mean the unref bit?
18:29 jiffin ndevos: i thought it is refered in __posix_acl_get()
18:30 ndevos jiffin: yeah, and then it gets assigned to the ctx, and an unref is done?
18:30 * ndevos needs to check taht again
18:31 ndevos jiffin: indeed, __posix_acl_get() would do the ref++, but when something keeps a refenence, it should not ref-- it
18:31 glusterbot ndevos: ref's karma is now 1
18:31 glusterbot ndevos: ref's karma is now 0
18:32 jiffin ndevos: i use the same procedure in posix_acl_set
18:32 ndevos oh, THIS is fun! maybe __posix_acl_set_specific() should ref++ it too, and then it needs t oref-- at one point
18:32 glusterbot ndevos: ref's karma is now 1
18:32 glusterbot ndevos: oref's karma is now -1
18:33 ndevos ah, typos!
18:33 hagarth ndevos: this is how a ref leak happens ;)
18:33 ndevos hagarth: eEXACTLY
18:33 jiffin ndevos: may be we need to send a fix there too
18:34 ndevos jiffin: hmm, maybe, I remember something about out-of-memory and acls....
18:35 jiffin hagarth: that's why niels is working on new ref counting mechanism, may be he can check those scenarios in  there too
18:35 jiffin ndevos: sounds interesting
18:37 ndevos jiffin: ah, wait, I think your change is correct on the ref counting part, you *do* have the manual ref++ in there
18:37 glusterbot ndevos: ref's karma is now 2
18:37 ndevos the different ways of doing reference counting is so confusing...
18:37 * ndevos checks again
18:38 * jiffin too
18:39 Gaurav joined #gluster-dev
18:39 ndevos jiffin: the way the refcounting is done in posix-acl is just weird, the get() functions do not ref++ and posix_acl_ref() is almost never called
18:39 glusterbot ndevos: ref's karma is now 3
18:40 jiffin ndevos: hmm
18:40 jiffin so unrefering after the get is not required??
18:41 ndevos jiffin: well, it is currently not done in a consistent way, but it is not wrong either
18:42 ndevos I think I found a place to do the next refcounting cleanups
18:42 jiffin ndevos: :D
18:44 * ndevos his network is slow... new installations over the (long-distance) wifi isnt helping
18:44 * jiffin finds application for refcount where he touches in the gluster code
18:44 ndevos jiffin: so, the refcount looks ok, no need to change thatt
18:44 jiffin ndevos: thanks
18:45 ndevos jiffin: but the coments in handling_other_acl_related_xattr are still valid, I think
18:45 jiffin ndevos: i also liked ur idea of using sys_lgetxattr once and then use GF_REALLOC
18:45 ndevos jiffin: and that function could have a nicer name ;-)
18:46 ndevos people always complain that getxattr() is so slow, so lets try to reduce that whwre we can
18:46 jiffin but in that when i use allocate  what should be size of buffer
18:46 jiffin ?
18:47 jiffin ndevos: i just want to make sensible name
18:47 ndevos what is the maximum acl size?
18:48 ndevos the larger allocation and reducing the size would be faster than 2x getxattr, I think
18:48 jiffin ndevos: it should be maximum possible size of xattr
18:48 ndevos jiffin: yes, I think so too
18:48 jiffin acl size may vary depending on user and group no
18:49 jiffin should i use 256?
18:49 * ndevos looks for a #define that makes some sense
18:49 jiffin i may not sure that
18:50 ndevos maybe xlators/mount/fuse/src/fuse-bridge.h:#define GLUSTERFS_XATTR_LEN_MAX  65536
18:51 ndevos 256 is rather small, users do have many groups and all that can be put in an acl
18:51 jiffin and 65536 is very large
18:51 hagarth ref-- , oref++ for ref parity
18:51 glusterbot hagarth: oref's karma is now 0
18:51 glusterbot hagarth: ref's karma is now 2
18:51 ndevos yes, I thought it would be the limit of the page-size, but maybe not
18:52 hagarth ref--, ref--
18:52 glusterbot hagarth: ref's karma is now 1
18:52 glusterbot hagarth: ref's karma is now 0
18:52 hagarth all is well with the world now
18:52 ndevos now we're not allowed to use ref anymore?
18:52 ndevos thank you for correcting that!
18:52 jiffin hagarth++
18:52 glusterbot jiffin: hagarth's karma is now 69
18:53 ndevos oh, hagarth is reference counted too!
18:54 hagarth use ref but with all the care it needs
18:55 ndevos jiffin: I'm not sure how big an xattr can be... 4096 would be a good start, and maybe add a warning when getxattr returns 4096 bytes?
18:55 jiffin ndevos: we can use refplusplus or refminusminus
18:55 ndevos LOL
18:56 ndevos or just ++ref and --ref ?
18:56 jiffin ndevos: k
18:56 ndevos jiffin: do set that 4096 in a #define somewhere :)
18:57 jiffin ndevos: sure
18:57 jiffin ndevos: luckily glusterbot only knows postincrementing
18:58 ndevos yes, but I think hagarth is tracking count
18:59 jiffin ndevos++
18:59 glusterbot jiffin: ndevos's karma is now 171
18:59 ndevos jiffin++ :)
18:59 glusterbot ndevos: jiffin's karma is now 8
18:59 * hagarth has all the karma counts in his mind
18:59 jiffin ndevos: :)
19:07 ndevos hagarth: btw, did you see my email about the maintainer stats, and did you check the graphs?
19:12 hagarth ndevos: where was this?
19:12 ndevos hagarth: maintainers@gluster.org
19:13 * ndevos checks
19:13 ndevos hagarth: http://article.gmane.org/gmane.comp​.file-systems.gluster.maintainers/7
19:14 msvbhat hagarth: BTW, did you get time to check out the geo-rep failback thing?
19:15 * msvbhat is curious as to what happenned
19:15 hagarth msvbhat: not yet
19:15 hagarth ndevos: this looks cool. I had read the email but had not clicked on the links.
19:15 msvbhat hagarth: Ah, Okay... Maybe I'll try and do it myself tomrrow
19:16 ndevos hagarth: yeah, and it is not too difficult to create graphs either
19:16 hagarth ndevos: neat, ELK++ :D
19:16 glusterbot hagarth: ELK's karma is now 1
19:17 ndevos oh, indeed, ELK++ !
19:17 glusterbot ndevos: ELK's karma is now 2
19:17 ndevos I started with stats for the (failed) regression test logs, see what message-id gets loged most and such
19:18 shaunm_ joined #gluster-dev
19:18 ndevos but that needs some more tuning before it is useful, the Gerrit data is much nicer formatted
19:20 hagarth ndevos: yeah
19:21 * ndevos also has a logstash sticker on his laptop now ;-)
19:22 hagarth ndevos: cool !
19:25 ndevos hagarth: if you pass a username and crypt() hash, I'll add an account for you and you can click around yourself
19:27 dlambrig_ joined #gluster-dev
19:29 hagarth ndevos: mailing it over
19:43 ndevos hagarth: wow, that took a while to arrive! but now you can log in on http://elk.cloud.gluster.org
19:44 ndevos hagarth: the elk needs manually feeding of the gerrit data, so it is a couple of days old, but for trying it should do
19:45 hagarth ndevos: auth failures
19:45 ndevos hmm, lets see
19:46 ndevos hagarth: try again?
19:47 ndevos oh, I think the elastic search needs more memory :-/
19:48 ndevos hagarth: I've disabled the automatic regression test logs, and it seems to be up and running again
19:48 hagarth ndevos: works now
19:49 ndevos hagarth: okay, click on "dashboard" and then the "open folder" icon on the righ
19:49 ndevos t
19:50 ndevos that will give you 3 dashboard options, 2 for maintainers, and one for the regression-logs
19:50 hagarth ndevos: nice!
19:51 ndevos hagarth: the "open folder" icon can be used in the "visulize" menu too, that is where the single graphs are located
19:51 ndevos hagarth: I'm not sure if the stats are really correct, I dont know the actual usage of some of the gerrit fieds
19:52 ndevos hagarth: but, feel free to click around, create new diagrams (SAVE THEM!!!) and build dashboards
19:52 hagarth ndevos: yeah, will do that
19:52 ndevos hagarth: save things when you make something, it is not save automatically and you can easily loose your things :-
19:53 ndevos :-/
19:55 dlambrig_ joined #gluster-dev
19:56 dlambrig_ left #gluster-dev
20:03 ndevos hagarth: I'm leaving now, enjoy your day!
20:03 hagarth ndevos: thanks, good night!
20:16 obnox joined #gluster-dev
21:06 shaunm_ joined #gluster-dev
21:51 badone joined #gluster-dev
23:07 nishanth joined #gluster-dev
23:26 jrm16020 joined #gluster-dev
23:50 an joined #gluster-dev

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