Perl 6 - the future is here, just unevenly distributed

IRC log for #gluster-dev, 2017-02-24

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

All times shown according to UTC.

Time Nick Message
00:43 ankitr joined #gluster-dev
02:17 ankitr joined #gluster-dev
02:48 ilbot3 joined #gluster-dev
02:48 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/
03:20 ankitr joined #gluster-dev
03:29 Shu6h3ndu joined #gluster-dev
03:39 vimal joined #gluster-dev
03:44 sanoj joined #gluster-dev
03:53 itisravi joined #gluster-dev
03:56 magrawal joined #gluster-dev
03:57 prasanth joined #gluster-dev
04:02 Shu6h3ndu joined #gluster-dev
04:14 vimal joined #gluster-dev
04:16 gyadav joined #gluster-dev
04:29 atinm joined #gluster-dev
04:29 skumar joined #gluster-dev
04:36 ankitr joined #gluster-dev
04:36 nbalacha joined #gluster-dev
04:49 kdhananjay joined #gluster-dev
04:51 karthik_us joined #gluster-dev
05:04 nbalacha joined #gluster-dev
05:05 BlackoutWNCT joined #gluster-dev
05:08 BatS9 joined #gluster-dev
05:10 ndarshan joined #gluster-dev
05:22 skumar_ joined #gluster-dev
05:33 rafi joined #gluster-dev
05:41 skoduri joined #gluster-dev
05:42 apandey joined #gluster-dev
05:43 riyas joined #gluster-dev
05:48 skoduri_ joined #gluster-dev
05:58 pranithk1 joined #gluster-dev
06:00 apandey_ joined #gluster-dev
06:00 rastar joined #gluster-dev
06:19 sanoj joined #gluster-dev
06:20 hgowtham joined #gluster-dev
06:22 susant joined #gluster-dev
06:22 Humble joined #gluster-dev
06:28 ankitr joined #gluster-dev
06:30 rafi1 joined #gluster-dev
06:30 ankitr_ joined #gluster-dev
06:30 rejy joined #gluster-dev
06:33 nbalacha joined #gluster-dev
06:38 nthomas joined #gluster-dev
06:41 susant joined #gluster-dev
07:10 aravindavk joined #gluster-dev
07:25 msvbhat joined #gluster-dev
07:28 ankitr joined #gluster-dev
07:37 pkalever joined #gluster-dev
07:45 kdhananjay joined #gluster-dev
07:48 rafi1 joined #gluster-dev
07:56 nbalacha joined #gluster-dev
08:07 skumar_ joined #gluster-dev
08:09 skumar_ joined #gluster-dev
08:13 skumar__ joined #gluster-dev
08:23 apandey joined #gluster-dev
09:02 ashiq joined #gluster-dev
09:13 susant joined #gluster-dev
09:16 rastar joined #gluster-dev
09:41 pranithk1 xavih: apandey: Hey, I don't think I understand the problem completely about the metadata version checks. It shouldn't be inside that function is my understanding...
09:42 xavih pranithk1: why not ? if versions mismatch, we need to heal them for sure, even if other things are completely equal
09:43 pranithk1 xavih: It does heal in the sense it only adjusts the versions
09:43 xavih pranithk1: it doesn't heal if the only difference is the version itself
09:44 pranithk1 xavih: okay what is the meaning of heal when you say it? :-)
09:45 pranithk1 xavih: I mean, it doesn't need to heal any attributes. It just needs to adjust the versions alone
09:45 pranithk1 xavih: it does adjust these versions
09:45 pranithk1 xavih: I gave an example in the comment as a reply to your comment
09:46 * xavih is looking at the code
09:48 xavih pranithk1: I think I see what you are telling...
09:51 xavih pranithk1: yes, you are right. Versions will be healed. Sorry for the noise... :-/
09:52 xavih pranithk1: I didn't look deeper into the code after the find direction return...
09:53 pranithk1 xavih: Just metadata is a bit different, I feel it is better this way. i.e. not trusting the versions...
09:54 xavih pranithk1: it's ok for me
09:54 pranithk1 xavih: cool.
09:54 pranithk1 xavih: 3.10 is going out without this patch :-(
09:55 xavih pranithk1: oops :(
09:55 pranithk1 xavih: yeah they made the build
09:56 xavih pranithk1: I thought it was intended for Feb 28...
09:56 pranithk1 xavih: me too
09:56 pranithk1 xavih: I did see a build
09:59 pranithk1 xavih: I asked for a re-spin, let's see
10:05 apandey pranithk1: Sorry, was away for tea..
10:05 apandey pranithk1: I think version should be considered..
10:05 pranithk1 apandey: hmm... why?
10:06 apandey pranithk1: I mean in heal_find_directions
10:06 apandey pranithk1: It is working now because something is wrong in calculating sources.
10:07 pranithk1 apandey: just metadata heal is a bit different. The thing is even when versions don't match but the attributes are fine, we should just set all versions to the largest one
10:07 apandey pranithk1: Yeh, but it will not go to adjust version if healed_sink is not set and number of sources are lesss then ec->nodes.
10:08 apandey pranithk1: which is happening in the bug which I raised.
10:09 apandey pranithk1: follow the steps in https://bugzilla.redhat.com/show_bug.cgi?id=1425703 and you will see the issue.
10:09 glusterbot Bug 1425703: unspecified, unspecified, ---, sheggodu, ASSIGNED , [Disperse] Metadata version is not healing when a brick is down
10:12 pranithk1 apandey: let me test that once
10:15 apandey pranithk1: ok
10:20 skumar_ joined #gluster-dev
10:20 nbalacha joined #gluster-dev
10:23 ShwethaHP joined #gluster-dev
10:27 pranithk1 xavih: apandey: Do you guys think we can complete the patch https://review.gluster.org/16298 by Monday ish?
10:28 pranithk1 xavih: apandey: I need to update my management in half hour in a meeting. I didn't get a chance to review it yet :-(
10:28 pranithk1 xavih: apandey: After my patch is done, even I will be helpful there :-)
10:30 pranithk1 xavih: apandey: If you think it may take more time, just let me know. I need to give a probable timeline for this patch
10:30 apandey pranithk1: I have not gone through Xavi's recent comments. Will let you know once I go though it.
10:30 pranithk1 apandey: okay
10:31 apandey xavih : Do you think it is going in the direction of your previous comments and it is just the implementation where you have comments.?
10:33 msvbhat joined #gluster-dev
10:33 xavih pranithk1: it's a complex patch. We are touching delicate things. I think we need to change what I've already commented and I'll need a full review after that. Not sure how much time I will be able to spend on it this weekend. However I can reserve Monday to work on it if I haven't had enough time before
10:34 xavih apandey: yes, I think this is the right direction, though I need to verify that things get modified when they can and should. That's the complex part...
10:36 apandey xavih: I agree.
10:36 pranithk1 xavih: Please spend time with family over the weekend. I will be spending time on it during the weekend. You have already taken enough burden for 3.10 with me barely contributing. I will make it up :-)
10:38 xavih pranithk1: for me https://review.gluster.org/16468 is already ok. Only remain minor changes that can be delayed. Only apandey's issue needs to be checked. Otherwise it's a +1 from me
10:39 xavih pranithk1: don't worry, I'll try to review it anyway :)
10:39 pranithk1 apandey: I don't think that issue is introduced with this patch, so may be we can take it as a separate patch? Since sunil is anyway working on it, we can finish up this patch and we can all concentrate on your patch?
10:39 pranithk1 apandey: I was talking about the metadata issue you talked about.
10:42 apandey pranithk1: yes, That is ok.
10:43 pranithk1 apandey: I will send out the updated patch after my final testing. So may be by six(have some meetings)
10:44 pranithk1 xavih: ^^
10:44 xavih pranithk1: oh, sorry :P
10:45 xavih pranithk1: I'll make the final review on it :)
10:45 pranithk1 xavih: I mean in 2 hours. 6IST
10:45 xavih pranithk1: no problem
10:46 pranithk1 xavih: It is only adding static and the comment. But I wanted to test (I was telling you yesterday right :-) )
10:50 skumar joined #gluster-dev
11:18 atinm joined #gluster-dev
11:43 vbellur joined #gluster-dev
11:53 msvbhat joined #gluster-dev
11:59 susant joined #gluster-dev
12:00 pranithk1 xavih: apandey: I think I found one more bug in my testing...
12:00 apandey pranithk1: :)
12:00 pranithk1 xavih: apandey: ec_get_heal_info()
12:01 pranithk1 xavih: apandey: In error cases it does 'goto out' which doesn't populate anything in the dict
12:01 apandey pranithk1: in your patch ?
12:02 pranithk1 apandey: no for a long time
12:02 pranithk1 apandey: just check the function
12:03 xavih pranithk1: is it needed to fill dict ?
12:03 pranithk1 xavih: I think it should fill "heal"
12:03 xavih pranithk1: it will return an error, so does matter the contents of the dict ?
12:05 pranithk1 xavih: Hmm... I think it could be a bug in glfs-heal infra itself...
12:05 xavih pranithk1: will the caller of the getxattr check dict contents if getxattr itself returns an error ?
12:05 pranithk1 xavih: glfsh_process_entries()
12:05 pranithk1 xavih: line 648
12:06 apandey :vs
12:07 pranithk1 xavih: apandey: I think it is bug in glfsheal itself
12:07 pranithk1 xavih: apandey: Will address it separately
12:07 pranithk1 xavih: apandey: I am almost done with my testing. Finding some other bugs but not in the patch we reviewed so far :-)
12:15 pranithk1 xavih: apandey: Found a bug!
12:16 xavih pranithk1: in heal ?
12:16 apandey pranithk1: ??
12:17 pranithk1 xavih: apandey: Updated the patch with the comment. It is so embarrassing :-(
12:18 xavih pranithk1: I didn't see it...
12:19 pranithk1 xavih: Oh the comment is not shown?
12:19 pranithk1 xavih: No I see that it is published.
12:19 pranithk1 xavih: Are you saying you missed the bug?
12:19 xavih pranithk1: yes, yes. I'm saying that I missed that one in the review... sorry...
12:19 pranithk1 xavih: ah! got it. No need to say sorry. These are all part of the game :-)
12:21 apandey pranithk1: This is major one, always true. yeh, we should have picked this  :(
12:23 atinm joined #gluster-dev
13:07 pranithk1 xavih: I am not able to come up with a case where the file/directory will need only data heal or entry heal. Every time we modify something both metadata and data parts are getting modified
13:07 pranithk1 xavih: I am thinking it is probably better to modify the xattrs directly on the backed to make sure we test heal info where it only needs data heal but it still prints it in heal info
13:08 pranithk1 xavih: To prevent the kind of bug we missed in the review... this is the only idea I have. Any better way?
13:08 xavih pranithk1: no. I think there's no case where we only modify data. All data changes also cause metadata changes
13:09 pranithk1 xavih: okay I will do the xattr modification in that case.
13:13 nthomas joined #gluster-dev
13:15 ira joined #gluster-dev
13:45 pranithk1 xavih: Found one more good bug :-)
13:45 xavih pranithk1: another one ? :|
13:45 pkalever left #gluster-dev
13:46 pranithk1 xavih: it is difficult to catch it
13:46 pranithk1 xavih: all these find_direction functions do ec_dict_del_array (dict, EC_XATTR_VERSION..)
13:47 pranithk1 xavih: so once metadata one is over, version and dirty won't be present in the dict
13:47 xavih pranithk1: true
13:47 pranithk1 xavih: good bug no :-)
13:48 xavih pranithk1: but this should cause a lot of troubles...
13:48 pranithk1 xavih: no no
13:48 xavih pranithk1: where do we use the dict twice ?
13:48 pranithk1 xavih: It only happens for heal info
13:48 pranithk1 xavih: Nowhere else :-)
13:49 riyas joined #gluster-dev
13:51 pranithk1 xavih: I am thinking of copying the dictionary, it is not clean but I can't think of any other way...
13:51 xavih pranithk1: I don't find where it's used twice
13:51 pranithk1 xavih: no where else
13:51 pranithk1 xavih: all heals do either lookup/xattrop+fstat inside locks
13:51 pranithk1 xavih: so it always uses it just once :-)
13:52 pranithk1 xavih: Only now it will have to use it twice
13:53 xavih pranithk1: should it make sense to implement an ec_dict_get_version() ?
13:53 xavih pranithk1: ec_dict_get_array(), sorry
13:53 pranithk1 xavih: I am also thinking that one :-)
13:53 pranithk1 xavih: That is cleaner
13:53 xavih pranithk1: yes
13:54 pranithk1 xavih: Let me do that one
13:54 xavih pranithk1: ok
13:55 xavih pranithk1: I've to leave now. I'll check later and review the patch
13:55 pranithk1 xavih: cool cya
13:55 vbellur1 joined #gluster-dev
13:57 vbellur joined #gluster-dev
13:58 vbellur1 joined #gluster-dev
13:59 vbellur joined #gluster-dev
13:59 rastar joined #gluster-dev
13:59 vbellur1 joined #gluster-dev
14:00 vbellur joined #gluster-dev
14:01 vbellur1 joined #gluster-dev
14:15 shyam joined #gluster-dev
14:16 rraja joined #gluster-dev
14:21 nbalacha joined #gluster-dev
14:39 ankitr joined #gluster-dev
14:55 shyam misc: Any reason you are not marking your commits +1 verified (other than the possible fact that you have not verified them)?
14:55 misc shyam: I didn't knew I could
14:55 misc Nigel told me this morning that I should
14:55 susant left #gluster-dev
14:56 misc but to me, verified was "someone else verified the commit to have it run on jenkins"
14:56 shyam misc: Oh, ok... cool, so you know without that and without the regression scores we cannot merge them ;)
14:56 misc ie, avoid someone putting some crappy code running on jenkins without oversight
14:56 shyam Also, as a practice we review once that verification is set, so we know things work somewhere... :)
14:56 misc (like, people committing code that mine coin in PR on github...)
14:57 shyam Hmmm...
14:57 misc and then, I had expenses to do :)
14:57 misc but I will add verified
14:57 shyam ok...
14:58 misc but so, for verified, I do need to run gluster and test the commit, or just verify it is the one I want ?
14:58 shyam Well... you verified your change is good.
14:59 misc oki, yeah, i can do that
14:59 shyam Now, ideally that involves (at least) a build
14:59 shyam and most often running some test that exercises the code
15:00 misc isn't the goal of smoke test of doing a build ?
15:01 shyam Yes, but we do not want the build to fail in the Jenkins systems, we would rather it not fail when someone submits the code BUT... I could be repeating something from history here
15:01 shyam I mean it is surely bad style to submit code that fails to compile and link
15:02 misc yep
15:12 rastar joined #gluster-dev
15:47 lpabon joined #gluster-dev
16:00 ankitr joined #gluster-dev
16:06 wushudoin joined #gluster-dev
16:09 jobewan joined #gluster-dev
16:16 pranithk1 joined #gluster-dev
16:53 ankitr joined #gluster-dev
17:03 rastar joined #gluster-dev
17:04 misc shyam: since you tagged 3.10, can I reboot stuff this weekend ?
17:04 misc (I figured that maybe, I should ask)
17:05 misc (or should I try to avoid rebooting jenkins and gerrit, and focus on the others ?)
17:28 major joined #gluster-dev
17:40 shyam misc: yes, you could, no blockers for me to handle
17:41 misc oki, I should have asked sooner, but I forgot, and we have kernel packages since 2 days
17:41 misc and the guy has a exploit to be published soon
17:41 shyam Let's update, we should be fine.
17:42 ankitr joined #gluster-dev
17:57 vbellur joined #gluster-dev
18:44 shyam joined #gluster-dev
19:17 shyam joined #gluster-dev
20:13 lkoranda joined #gluster-dev
20:14 vbellur joined #gluster-dev
20:20 csaba joined #gluster-dev
20:26 lkoranda joined #gluster-dev
21:00 vbellur joined #gluster-dev
21:08 JoeJulian Oh, wow. Dan Lambright's left Red Hat, huh? Does anyone know if he's still going to work on gluster? The dedup work he was working on seemed really interesting.
21:25 Acinonyx joined #gluster-dev
21:46 Acinonyx joined #gluster-dev
22:05 Acinonyx joined #gluster-dev
22:11 Acinonyx joined #gluster-dev

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