Camelia, the Perl 6 bug

IRC log for #gluster-dev, 2013-04-08

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

All times shown according to UTC.

Time Nick Message
00:25 yinyin joined #gluster-dev
01:14 jules_ joined #gluster-dev
01:34 jules_ joined #gluster-dev
01:42 precipitousquarr joined #gluster-dev
01:43 precipitousquarr Hi do the blocks need to be at the root of a file system? ex. can I have /mnt/d1/subfolder1 be one block and /mnt/d1/subfolder2 be another block?
01:58 bala1 joined #gluster-dev
02:47 precipitousquarr joined #gluster-dev
02:50 vshankar joined #gluster-dev
02:54 hagarth JoeJulian: #gluster has gone invite only again. Can you please check if glusterbot can do -i there?
03:40 johnmark doh
03:48 sgowda joined #gluster-dev
04:06 sripathi joined #gluster-dev
04:11 bulde joined #gluster-dev
04:21 yinyin joined #gluster-dev
04:26 mohankumar joined #gluster-dev
04:32 anmol joined #gluster-dev
04:37 hagarth joined #gluster-dev
05:01 puebele joined #gluster-dev
05:02 yinyin joined #gluster-dev
05:15 sgowda joined #gluster-dev
05:21 aravindavk joined #gluster-dev
05:27 deepakcs joined #gluster-dev
05:49 rastar joined #gluster-dev
05:54 raghu joined #gluster-dev
06:07 sgowda joined #gluster-dev
06:39 bala1 joined #gluster-dev
06:44 yinyin joined #gluster-dev
06:47 sgowda joined #gluster-dev
06:49 hagarth joined #gluster-dev
06:54 sripathi joined #gluster-dev
06:54 ollivera joined #gluster-dev
07:34 sgowda joined #gluster-dev
07:53 hagarth joined #gluster-dev
08:25 lalatenduM joined #gluster-dev
09:02 rastar joined #gluster-dev
09:06 sripathi joined #gluster-dev
09:10 rastar1 joined #gluster-dev
09:51 edward1 joined #gluster-dev
10:13 bulde1 joined #gluster-dev
10:18 rastar joined #gluster-dev
10:38 precipitousquarr joined #gluster-dev
11:11 hagarth joined #gluster-dev
12:04 ollivera Hi there
12:04 ollivera I would like to hash a file name so that it end up in a brick that matches that hash range.
12:34 yinyin joined #gluster-dev
12:46 hagarth joined #gluster-dev
12:59 bulde joined #gluster-dev
13:34 ollivera Does the hashing happen on the file name or the complete file path?
13:37 kkeithley The hashing is only on the filename
13:43 ollivera kkeithley, do you know how the hashing algorithm works? I want to rename the file so that it should end up in other "server"
13:43 ndevos joined #gluster-dev
13:46 kkeithley libglusterfs/src/hashfn.c, gf_dm_hashfn() is what the dht xlator calls to determine which brick to place files on.
13:47 kkeithley There's also an undocumented way to force placement that's probably easier to use than trying to coerce a particular hash
13:49 ollivera kkeithley, I looked at that function. however, how do I do the opposite? From a layout, get the rule to give a name to a file.
13:49 ollivera kkeithley, where is it?
13:49 kkeithley where is what?
13:50 ollivera kkeithley"undocumented way to force ..."
14:01 ollivera kkeithley, my "c" knowledge doesn't enable me to do reverse engineering in gf_dm_hasfn()
14:01 ollivera :)
14:12 wushudoin joined #gluster-dev
14:13 kkeithley I don't remember what the magic undocumented method is. I just know it's in there. JoeJulian, is there a glusterbot hint with this? Probably not. I did a quick google search but didn't come up with anything.
14:14 kkeithley And it's undocumented because we don't actually recommend that people use it, (never mind that I seem to have recommended it. ;-))
14:24 lalatenduM joined #gluster-dev
14:32 ollivera kkeithley, it is quite hard to check the source code when you are not familiar with. Looking for a function dht_create_filename_for_layout () :)
14:33 jbrooks joined #gluster-dev
14:40 rastar joined #gluster-dev
14:54 ndevos ollivera: what would you expect to see as a return value for a dht_create_filename_for_layout() function? A(n endless) list of all possible filenames that would go to a specific subvolume?
14:59 ollivera ndevos, yeah. it doesn't make sense. I just want a function that returns a filename that fits in that layout
15:03 ollivera ndevos, does it make sense?
15:04 ndevos ollivera: yeah, that makes more sense
15:04 ollivera ndevos, is that possible? does gluster provide such function?
15:06 ndevos ollivera: not that I know
15:06 ndevos ollivera: but, a force placement like kaleb mentioned can be done: http://lists.nongnu.org/archive/html​/gluster-devel/2013-04/msg00013.html
15:07 kkeithley I knew someone would be able to find that.
15:26 copec left #gluster-dev
15:44 bulde joined #gluster-dev
16:00 johnmark portante: ping
16:25 johnmark hagarth: ping
16:26 johnmark kkeithley: ping
16:27 johnmark can either of you look into http://review.gluster.org/#/c/4689/ - for backporting to 3.3?
16:31 johnmark wait, I guess it's already backported, just need a release schedule for 3.3.2
16:40 johnmark and a confirmation that the bug is actually resolved - https://bugzilla.redhat.com/show_bug.cgi?id=922809
16:40 glusterbot Bug 922809: unspecified, unspecified, ---, pkarampu, POST , Rename fails in dbench when a brick is taken down.
16:41 JoeJulian @later tell ollivera wrt dht hashing (or any hashing function): All hashing functions are lossy, so no. There's no way to take a hash and determine the filename. You could probably reverse-engineer a filename if you made assumptions, but I haven't seen an algorithm to do that.
16:41 glusterbot JoeJulian: The operation succeeded.
16:42 semiosis JoeJulian: rainbow tables
16:43 JoeJulian I suppose you could build a rainbow table with every possible filename and select a list of filenames that would match the hash you want, but that seems pretty overkill.
16:52 johnmark semiosis: rainbow of fruit flavors? what?
16:52 semiosis taste the rainbow table
16:53 johnmark heh
17:14 kkeithley johnmark: wrt  http://review.gluster.org/#/c/4689/ I see no evidence that it was ever reviewed or merged into the tree.
17:15 kkeithley Looking at the source confirms it's never been accepted or merged.
17:19 kkeithley You should check with Pranith to see what happened
17:19 johnmark kkeithley: gotcha - thanks
17:20 johnmark kkeithley: is he ever in the channel?
17:48 semiosis @seen pranithk
17:48 glusterbot semiosis: pranithk was last seen in #gluster-dev 2 weeks, 3 days, 3 hours, 54 minutes, and 24 seconds ago: <pranithk> xavih: One request... I went through the code. Could you please give getfattr output on that problematic directory when you get to it. If I am not around please add output of "getfattr -d -m . -e hex <dir>" to the bug I mentioned above..
17:53 H__ Are there known replace-brick issues ? I found none in my search yet see replace-brick hang in two setups, both on 3.3.1 (see also todays mail to gluster users)
18:14 portante johnmark: pong
18:31 hagarth johnmark: ping
19:05 johnmark portante: need UFO on the forge, pronto - start wiht a couple of docs, and whatever repo you're using at the moment
19:06 johnmark hagarth: you already responded to my earlier request
19:07 hagarth johnmark: noticed that later
19:07 johnmark hagarth: no worries - don't you need to sleep or something???
19:08 hagarth johnmark: will head soon - to be at least awake in time for the glusterd call :)
19:11 portante johnmark: right the scaling work is in downstream rhs repos, internal to redhat only
19:11 portante junaid has started porting these changes upstream to the glusterfs upstream repo
19:12 johnmark portante: gah
19:12 johnmark portante: whose idea was that?
19:12 portante johnmark: customer fire fight
19:12 johnmark portante: oh oh right, ok
19:12 johnmark now I remember
19:12 portante we had to move quickly to solve their problem
19:12 johnmark yup, I remember the story
19:12 portante crazy times
19:12 johnmark portante: will ping junaid
19:13 portante k
19:13 portante what is the URL again for the forge?
19:13 johnmark portante: forge.gluster.org
19:13 johnmark kkeithley: can you help with this?
19:13 portante we gotta get that certificate fixed
19:13 johnmark portante: it's on order - we just need our IT guys to respond to the request :)
19:14 portante k
19:27 kkeithley help with what?
19:37 johnmark kkeithley: UFO on forge
19:42 kkeithley installing it? What's it running, CentOS? just get the rpms from my fedorapeople.org repo
19:53 johnmark kkeithley: docs, repo - if you want to know "what is UFO" "how to install it" and "what do I need to get it working" - there needs to be a project page for it
20:13 kkeithley okay, I added a card to the ToDo column of the UFO board in Trello
20:13 portante joined #gluster-dev
20:54 jbrooks joined #gluster-dev
21:41 badone What is the significance of the alloc-count value in a state dump? Is it memory allocated? Is alloc-count x padded_sizeof a reflection of the structures that have been allocated?
23:02 badone It looks like in mem_get() we do mem_pool->alloc_count++; so it looks like each time memory for a struct is allocated this count gets incrememtned so it looks like this is a count of the number of times memory has been allocated, not the same as what is curretnly allocated (hot_count?)

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