Perl 6 - the future is here, just unevenly distributed

IRC log for #gluster-dev, 2015-11-12

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

All times shown according to UTC.

Time Nick Message
00:30 hgichon joined #gluster-dev
00:35 shaunm joined #gluster-dev
01:03 zhangjn joined #gluster-dev
01:03 EinstCrazy joined #gluster-dev
01:34 mohan_ joined #gluster-dev
02:40 amye joined #gluster-dev
02:45 amye left #gluster-dev
02:49 mohan_ joined #gluster-dev
02:58 mohan joined #gluster-dev
03:06 overclk joined #gluster-dev
03:07 pranithk joined #gluster-dev
03:13 kdhananjay joined #gluster-dev
03:21 itisravi joined #gluster-dev
03:37 suliba joined #gluster-dev
03:39 kanagaraj joined #gluster-dev
03:58 sakshi joined #gluster-dev
04:00 mohan joined #gluster-dev
04:16 kanagaraj_ joined #gluster-dev
04:18 kdhananjay1 joined #gluster-dev
04:25 gem joined #gluster-dev
04:28 vimal joined #gluster-dev
04:32 kdhananjay joined #gluster-dev
04:40 ppai joined #gluster-dev
04:43 nishanth joined #gluster-dev
04:44 hgowtham joined #gluster-dev
04:48 itisravi sakshi: hey! when dht gets CHILD_MODIFIED, does dht explicitly send an internal lookup to read layout on all dirs? or does it send it when the mount performs a lookup?
04:48 sakshi itisravi: so when a C
04:49 atalur joined #gluster-dev
04:49 sakshi CHILD_MODIFIED event happens the conf->gen is incremented and will be greater that the layout->gen of the directories, hence when a lookup comes on them it will forget the old layout and get new layout
04:53 ndarshan joined #gluster-dev
04:53 itisravi sakshi: so in the case that you described, b0 goes down, mkdir happens, b0 comes back, but before lookup comes on it, b1 goes down. Will dht create the dir on bo and set the currect layout for it while b1 is still down?
04:56 sakshi itisravi, yes, since as long as even one of the replica pair is up, for DHT that subvol is up and running, it can create the dir on bo
04:57 itisravi sakshi: hmm ok..but if we did not increment conf->gen, then the dir creation on bo wouldn't happen even if lookup came from the mount is ut?
04:57 itisravi it*
04:58 mchangir_ joined #gluster-dev
05:01 sakshi itisravi, if we do not increment it then the fresh lookup wont happen, rather  a revalidate lookup happens that would not heal the directory and it wont be created on bo
05:01 itisravi sakshi: okay.
05:05 itisravi sakshi: So the GF_EVENT_SOME_CHILD_DOWN is an optimization no? The actual fix for the bug reported in the BZ is changing the ENOENTS to EIOs in dht I guess?
05:07 kdhananjay joined #gluster-dev
05:08 sakshi itisravi: the main idea of the patch is to add a new event GF_EVENT_SOME_CHILD_DOWN which would ensure that when a single replica pair comes down we do not increment the conf->gen and do not do fresh lookup's. The ENOENT to EIO change in DHT is a side change to display proper errors
05:09 kdhananjay joined #gluster-dev
05:09 aravindavk joined #gluster-dev
05:09 itisravi sakshi: right, so I was wondering if it makes sense to split it into 2 patches. If one looks at the BZ description and the commit message, things wouldn't seem to correlate..
05:10 sakshi itisravi, sure Il remove the ENONET to EIO changes from the patch for better clarification
05:10 itisravi cool!
05:10 itisravi sakshi++
05:10 glusterbot itisravi: sakshi's karma is now 14
05:13 asengupt joined #gluster-dev
05:14 itisravi sakshi: Maybe you can send the AFR changes against a separate BZ (For optimiszing when DHT can avoid fresh lookups).
05:15 sakshi itisravi, yes will do that
05:25 jvava joined #gluster-dev
05:26 jvava here we have a 3node gluster fs. question is why the load of both server and client get very high after starting self-heal, so i have to stop self-heal, but files will become inconsistent. please help.
05:26 itisravi jvava: There are improvements being done to self-heal performance in glusterfs 3.8. What kind of files are there in your volume?
05:30 josferna joined #gluster-dev
05:30 jvava many plain files, medium size.
05:32 itisravi_ joined #gluster-dev
05:32 jvava about 1 thouthends file, size about 15T bytes
05:34 itisravi_ jvava: 15TB/1000 means about 15GB average size of file.
05:34 itisravi_ jvava: You can change the self-heal algorithm to full to see if it makes a difference to the load.
05:35 itisravi_ jvava: aslo if you're using glusterfs 3.7. you can enable sharding. That should also speed up self-heal.
05:36 itisravi_ Atleast for any new files that are created.
05:41 jvava typos, not 1 thousand,  we have 50 milions of files, volume size 15t
05:43 jvava what can we do except for upgrading gluster fs from 3.5 to 3.7
05:51 ashiq joined #gluster-dev
05:55 ashiq joined #gluster-dev
06:02 hagarth joined #gluster-dev
06:13 jvava anyway, gime more clues, thank you
06:16 jvava about high load by self-heal
06:18 jvava here, gfs: 3.5, vs: 15t, fs: 50m, load 1, cpu: 1000%
06:19 jvava is it related to layout of directory, is said a directory should contain less files as possible
06:27 mchangir joined #gluster-dev
06:27 josferna joined #gluster-dev
06:32 josferna_ joined #gluster-dev
06:45 hgowtham joined #gluster-dev
06:57 spalai joined #gluster-dev
07:05 rafi joined #gluster-dev
07:10 josferna_ joined #gluster-dev
07:18 josferna__ joined #gluster-dev
07:23 skoduri joined #gluster-dev
07:29 kdhananjay joined #gluster-dev
07:29 zhangjn joined #gluster-dev
07:31 hgowtham joined #gluster-dev
07:33 zhangjn joined #gluster-dev
07:42 josferna_ joined #gluster-dev
07:45 josferna joined #gluster-dev
07:53 hgowtham joined #gluster-dev
08:05 ashiq joined #gluster-dev
08:06 mohan_ joined #gluster-dev
08:13 itisravi joined #gluster-dev
08:15 aravindavk joined #gluster-dev
08:21 josferna_ joined #gluster-dev
08:36 jvava because our gluster fs went wrong recently very often, boss demand us fix it and work on weekend, so, please help, or we will live without weekend, thanks.
08:40 Pablo|off| joined #gluster-dev
08:45 itisravi joined #gluster-dev
08:50 hgowtham joined #gluster-dev
09:26 hchiramm joined #gluster-dev
09:37 hagarth left #gluster-dev
09:43 hchiramm joined #gluster-dev
09:54 firemanxbr joined #gluster-dev
09:56 hchiramm joined #gluster-dev
10:26 hgowtham joined #gluster-dev
10:36 deepakcs joined #gluster-dev
10:48 lkoranda_ joined #gluster-dev
10:58 lkoranda joined #gluster-dev
11:09 ashiq joined #gluster-dev
11:25 rafi joined #gluster-dev
11:27 kkeithley1 joined #gluster-dev
11:33 shyam joined #gluster-dev
11:36 kdhananjay joined #gluster-dev
11:37 rafi joined #gluster-dev
11:39 lpabon joined #gluster-dev
11:41 bfoster joined #gluster-dev
11:58 amye joined #gluster-dev
12:04 rafi joined #gluster-dev
12:12 bfoster joined #gluster-dev
12:33 kdhananjay joined #gluster-dev
12:44 spalai left #gluster-dev
12:44 ira joined #gluster-dev
13:22 dlambrig_ joined #gluster-dev
13:35 shaunm joined #gluster-dev
13:39 ira joined #gluster-dev
13:47 overclk joined #gluster-dev
14:09 amye joined #gluster-dev
14:10 amye left #gluster-dev
14:27 pranithk joined #gluster-dev
14:54 mohan_ joined #gluster-dev
15:29 gem joined #gluster-dev
16:01 cholcombe joined #gluster-dev
16:21 kbyrne joined #gluster-dev
16:36 RedW joined #gluster-dev
16:39 firemanxbr joined #gluster-dev
17:08 gem joined #gluster-dev
17:15 amye joined #gluster-dev
17:16 amye1 joined #gluster-dev
18:53 ilbot3 joined #gluster-dev
18:53 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/
19:00 shaunm joined #gluster-dev
19:29 wushudoin joined #gluster-dev
19:45 cholcombe joined #gluster-dev
19:47 suliba joined #gluster-dev
19:47 cholcombe joined #gluster-dev
19:50 amye joined #gluster-dev
19:54 dlambrig_ joined #gluster-dev
19:58 cholcombe joined #gluster-dev
20:04 cholcombe joined #gluster-dev
20:10 dlambrig_ joined #gluster-dev
20:34 ilbot3 joined #gluster-dev
20:34 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/
20:39 shaunm joined #gluster-dev
20:54 amye joined #gluster-dev
21:20 ira joined #gluster-dev
21:29 dlambrig_ joined #gluster-dev
21:56 shyam joined #gluster-dev
22:13 mjrosenb has the way linkto works changed recently? (since 3.3 or so)
22:16 mjrosenb or is lookup/stat/etc. supposed to always return the information of the linkto file?
22:17 amye joined #gluster-dev
23:13 wushudoin joined #gluster-dev
23:15 wushudoin joined #gluster-dev
23:20 jtc joined #gluster-dev
23:49 hgichon joined #gluster-dev
23:57 sankarshan_away joined #gluster-dev

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