Wednesday, 2010-08-18

*** sophiap has joined #openstack00:02
*** sophiap has quit IRC00:20
*** ptremblett has quit IRC00:24
*** zheng_li has quit IRC00:26
*** syah has quit IRC00:27
vishyI'd like to reiterate00:34
vishynetworking is hard, lets go shopping00:34
*** syah has joined #openstack00:34
vishyis there any way to see the diff between 2 revisions on launchpad?00:35
gholtcw: Yeah, in theory you should be able to change the ring by a power of two and still have the data on the devices they need to be. But... You still need to move all that data to the new paths.00:38
gholtSo, in practice, it'd be a pain and helluva impact.00:38
zaitcevI guess you should be able to download and then... diff -urpN -X dontdiff a/ b/00:38
*** gustavomzw has joined #openstack00:38
zaitcevif everything else fails00:39
*** sandywalsh has joined #openstack00:41
gholtcw: Well, I change my mind. Maybe it wouldn't be that much of an impact. There aren't usually that many partitions on a device so maybe combining/splitting them wouldn't be too bad.00:44
gholtcw: That'd be some fun code to write and test. :)00:44
*** gustavomzw has quit IRC00:45
*** jero- has quit IRC00:48
*** maplebed has quit IRC00:53
*** jc_smith has quit IRC00:56
*** jdmaturen has quit IRC00:57
*** sirp1 has quit IRC01:02
*** Glace has quit IRC01:25
*** Glace has joined #openstack01:32
*** dendrobates is now known as dendro-afk01:34
*** rsampaio has joined #openstack01:36
*** rdw has quit IRC01:45
*** jero- has joined #openstack01:54
*** jdmaturen has joined #openstack01:54
*** sirp1 has joined #openstack01:58
jero-why is that not your name?02:04
*** abecc has quit IRC02:10
notmynamemaybe it is my name ;-)02:11
gholtHis parents were enamored with the Liar's Paradox.02:12
*** e1mer has quit IRC02:14
*** Xenith has quit IRC02:15
*** jdmaturen has quit IRC02:16
*** miclorb_ has quit IRC02:18
mtaylorvishy: yes02:19
jero-anyone here from the nova storage layer ? I'm trying to draw a map of the IO path between guest and storage + the path setup02:25
*** jamiew_moviez has quit IRC02:26
*** rsampaio has quit IRC02:33
*** mordred_ has joined #openstack02:35
*** ChanServ sets mode: +v mordred_02:35
*** mtaylor has quit IRC02:37
*** mordred_ is now known as mtaylor02:39
*** e1mer has joined #openstack02:48
*** almaisan-away is now known as al-maisan02:51
*** RobertLJ has joined #openstack02:52
*** e1mer has quit IRC02:53
*** rajijoom has joined #openstack02:55
*** al-maisan is now known as almaisan-away03:02
*** novas0x2a has joined #openstack03:02
*** sophiap has joined #openstack03:05
*** RobertLJ has quit IRC03:08
*** sophiap has quit IRC03:11
*** e1mer has joined #openstack03:20
*** sophiap has joined #openstack03:20
*** miclorb has joined #openstack03:40
*** mordred_ has joined #openstack03:42
*** ChanServ sets mode: +v mordred_03:42
*** jero-- has joined #openstack03:45
*** mtaylor has quit IRC03:46
*** jero- has quit IRC03:46
cwgholt: i'm going to add it to my TODO list to model at some point03:48
cwgholt: i think it's potentially something we might encounter here03:48
*** pag has joined #openstack04:01
*** pag has quit IRC04:03
*** sandywalsh has quit IRC04:08
*** Phacops has quit IRC04:12
*** Phacops has joined #openstack04:17
*** justinsheehy has quit IRC04:18
*** justinsheehy has joined #openstack04:24
*** miclorb has quit IRC04:30
*** miclorb has joined #openstack04:36
*** sirp1 has quit IRC04:41
*** kashyapc has joined #openstack04:42
*** zaitcev has quit IRC04:50
*** jdmaturen has joined #openstack05:09
*** jdmaturen has quit IRC05:10
*** silassewell has quit IRC05:19
*** almaisan-away is now known as al-maisan05:22
*** al-maisan has left #openstack05:23
*** guigui has joined #openstack05:28
*** Xenith has joined #openstack05:34
*** rdw has joined #openstack05:40
*** jsgotangco has joined #openstack05:49
*** lmcdowell has quit IRC06:26
*** lmcdowell has joined #openstack06:32
*** ibarrera has joined #openstack06:33
*** justinsheehy has quit IRC06:41
*** allsystemsarego has joined #openstack06:47
*** justinsheehy has joined #openstack06:47
*** brd_from_italy has joined #openstack06:48
*** jsgotangco has quit IRC06:49
*** perestrelka has quit IRC06:49
*** guigui has quit IRC06:59
*** e-vad has quit IRC07:03
*** e-vad has joined #openstack07:11
*** fsoyer has left #openstack07:20
*** kainam has joined #openstack07:24
*** kainam- has quit IRC07:24
*** zheng_li has joined #openstack07:45
*** calavera has joined #openstack08:03
*** miclorb has quit IRC08:08
*** Xenith has quit IRC08:09
*** jpopi has joined #openstack08:13
*** zheng_li has quit IRC08:17
*** brd_from_italy has quit IRC08:28
*** gasbakid has joined #openstack08:35
*** zheng_li has joined #openstack08:35
*** miclorb has joined #openstack08:42
sorentermie: I addressed your concerns in
*** comstud has quit IRC08:46
*** ChrisAM has quit IRC08:46
termiesoren: approvizzled08:48
*** ChrisAM has joined #openstack08:49
*** ChrisAM has quit IRC08:54
*** ChrisAM has joined #openstack08:56
*** ChrisAM has quit IRC09:03
*** sophiap has quit IRC09:04
*** ChrisAM has joined #openstack09:07
sorentermie: \o/09:14
*** perestrelka has joined #openstack09:14
*** rsampaio has joined #openstack09:19
*** miclorb has quit IRC09:20
*** novas0x2a has quit IRC09:27
*** novas0x2a has joined #openstack09:28
*** rsampaio has quit IRC09:53
*** rajijoom has quit IRC09:56
*** miclorb has joined #openstack09:57
*** rnewson has joined #openstack10:34
*** justinsheehy has quit IRC10:44
*** avsm has joined #openstack10:45
*** ibarrera has quit IRC10:45
*** calavera has quit IRC10:46
*** justinsheehy has joined #openstack10:56
*** calavera has joined #openstack10:59
*** ibarrera has joined #openstack11:01
*** Kami__ has joined #openstack11:03
*** Kami_ has quit IRC11:03
*** justinsheehy has quit IRC11:08
*** miclorb has quit IRC11:19
*** sandywalsh has joined #openstack11:29
*** ctennis has quit IRC11:36
*** miclorb_ has joined #openstack11:45
*** jero-- has quit IRC11:46
*** ctennis has joined #openstack11:54
*** justinsheehy has joined #openstack11:58
*** Podilarius has left #openstack11:59
*** londo__ has joined #openstack12:03
*** joschi has joined #openstack12:05
*** londo_ has quit IRC12:07
*** gundlach has joined #openstack12:11
*** chmouel has quit IRC12:15
*** avsm has quit IRC12:16
*** ambo_ has joined #openstack12:22
sorengundlach: Maybe your question in revision 244 of holds its own answer?12:25
* gundlach scrambles to figure out his question12:25
sorengundlach: Look at r244 on that page, and you'll see.12:26
gundlachah, yes! thanks!12:26
gundlachok, next question -- why not pull my description of change into the commit -- or the other way around?12:26
sorengundlach: That's what we're going to do from now on.12:26
sorenI'm not sure I like that, though.12:27
sorengundlach: For two reasons:12:27
*** abecc has joined #openstack12:27
*** kashyapc has quit IRC12:27
sorengundlach: a) What you put in the description may no longer be true once the branch has gone through review and maybe has been changed.12:27
sorengundlach: b) I consider them quite different. What I want in the commit log simply isn't necessarily the same as what I want to tell people who intend to review it.12:29
sorenb) is a matter of taste more than anything, though. I feel a) applies globally.12:29
gundlachsoren: i see on both points.  thanks.12:31
sorengundlach: sure thing.12:31
gundlachif we ever did decide to go back to having two messages, perhaps fixing the UI would help -- so that newbies to LP like me would understand the context in which the message was to be read.12:31
*** miclorb_ has quit IRC12:35
sorenYeah, there's clearly room for improvement there.12:37
*** lmcdowell has quit IRC12:38
sorengundlach: Do you think something simple like  "Enter a message for the reviewer here" and "Enter the desired commit message here" would have done the trick?12:42
gundlachsoren: Probably not, because people coming from another tool would think "OK, my commit message *is* my message for the reviewer"12:43
gundlachmaybe more like:12:43
gundlach"Enter your commit message"12:43
gundlach"If you want to leave a special note for the reviewer, do so here.  It will only be used during the merge process".12:44
gundlach(or wherever it's used; i already forget because i'm having trouble keeping them straight :) )12:44
sorenbug 61981012:45
uvirtbotLaunchpad bug 619810 in launchpad "Difference between description and commit message in merge proposal is unclear" [Undecided,New]
gundlachbut really, i think more people will be happy with a simpler UI that uses a single message.  LP is busy enough that collating ideas is probably a good thing :)12:45
*** lmcdowell has joined #openstack12:45
*** jdarcy has joined #openstack12:53
*** pvo has joined #openstack12:56
*** ChanServ sets mode: +v pvo12:56
*** ambo_ has quit IRC12:57
*** gustavomzw has joined #openstack12:57
*** lmcdowell has quit IRC13:03
*** lmcdowell has joined #openstack13:09
*** ambo_ has joined #openstack13:11
*** pvo has quit IRC13:14
*** ratasxy has joined #openstack13:23
*** gasbakid has quit IRC13:23
ratasxyhello, when it leaves the final version of openstack?13:25
gundlachratasxy: around october is the first release.13:26
gundlachit will not be the "final" version -- we will keep releasing new versions as we improve it.13:26
ratasxythanks gundlach13:26
*** gasbakid has joined #openstack13:26
ratasxygundlach,  exist a beta?13:27
*** f4m8 has joined #openstack13:27
gundlachratasxy: you can get the current code at
gundlachthere are two parts: "nova", which manages virtualized machines (like EC2), and "swift", which handles storage (like S3)13:28
*** Fido has joined #openstack13:29
zuldid the derootification make it it in today?13:31
*** guigui has joined #openstack13:35
*** pvo has joined #openstack13:39
*** ChanServ sets mode: +v pvo13:39
*** Fido has quit IRC13:40
*** rajijoom has joined #openstack13:46
*** sophiap has joined #openstack13:49
jerowhat does the derootification imply ?13:50
*** comstud has joined #openstack13:53
*** ChanServ sets mode: +v comstud13:53
sorenzul: Still needs one more positive review.13:55
sorenjero: Currently, everything runs as root.13:55
sorenjero: Well, that's not entirely true.13:55
sorenjero: they run as the user who starts them.13:55
sorenjero: ...but if you run from init scripts, that's root.13:55
sorenjero: The derootification branch lets the daemons start as root and drop privileges to a regular user.13:56
soren...and for the looks for it, it needs more work. Darn it.13:57
* soren is having a complete preposition fail today13:59
soren"/by/ the looks /of/ it"...14:00
jeroI thought it was related to some tree-based architecture14:00
sorenjero: Oh, no, nothing like that.14:00
jerosoren: is that for swift ?14:00
sorenjero: No, nova.14:00
jeroi definately need a coffee cup.14:01
*** kashyapc has joined #openstack14:07
*** Podilarius has joined #openstack14:11
*** hazmat has joined #openstack14:11
*** ppetraki has joined #openstack14:25
*** pcrews has joined #openstack14:28
*** mordred_ is now known as mtaylor14:37
*** sirp1 has joined #openstack14:42
zulhow do you create project network data?
*** ibarrera has quit IRC14:43
*** f4m8 is now known as f4m8_14:46
*** justinsheehy has quit IRC14:49
*** pcrews is now known as pcrews_bbl14:55
*** tr3buchet has joined #openstack14:56
*** aliguori has joined #openstack14:57
sorenzul: You create a project using "nova-manage project create"14:59
sorenzul: Whether that's what it refers to, I'm not sure.14:59
*** calavera has quit IRC15:07
*** kashyapc_ has joined #openstack15:11
*** kashyapc has quit IRC15:12
*** asksol has quit IRC15:14
uvirtbotNew bug: #619901 in swift "add debian package dependencies" [Undecided,New]
edaygundlach: so, does disable-msg not work for you? I'm using latest pylint that ships with lucid and disable= doesn't work for me, which is why I had to switch back15:22
gundlachhmm; i did pip install pylint into my venv and disable-msg didn't work wit 0.21.115:22
edayhm, yeah I'm on 0.1915:23
jpipeseday: me too15:23
*** mrayzenoss has joined #openstack15:23
*** mrayzenoss is now known as mray15:23
*** mray has joined #openstack15:23
edaythats strange they just broke disable-msg, you'd think it would just deprecate it15:23
jpipeseday: but I think you can re-enable disable-msg with some command...15:23
gundlachi'd be fine changing pip-requires to pylint==0.19 until Ubuntu M ships, and discarding my branch15:24
gundlacheday: i think it was deprecated for a while?15:24
* jpipes admits there is little documentation on the pylintrc file, though!15:24
gundlach(was having to read up on it to figure out what i was doing wrong)15:24
jpipesgundlach: yeah, there's not much reading though ;)15:24
gundlachjpipes: maybe i went into the source? can't remember.  also, i like your new handle :)15:25
edayjpipes: so, there is a way to do renable disable-msg in 0.21? that might be good for now to transition15:25
jpipesgundlach: heh, didn't even notice...must have had the Internets kick out last night...15:28
jpipeseday: I believe there is, but will have to check the sources...I'd recommend sticking with 0.19 for right now...we can do a mass sweep replace for s/disable-msg/disable/ in a single patch in the future15:29
jpipesgundlach: ^^15:29
jpipeseday: which would make it nicer for folks merging (see ewan's latest merge conflict stuff from style/pep fixes)15:29
gundlachjpipes: yeah, i did sed -i s/disable-msg/disable/g in my patch.15:30
gundlachjpipes: ok i'll update my patch to change us to 0.19 and change disable to disable-msg throughout.15:30
jpipesgundlach: yeah, I say we table that until Ubuntu M at least, and add 0.19 to the pip-requires.15:30
jpipesgundlach: cheers15:30
edayjpipes, gundlach: cool, sounds good15:31
gundlachjpipes: where are you from? i don't here 'cheers' except from the irish15:31
jpipesgundlach: I'm British15:31
jpipesgundlach: but I've spent most of my life in the U.S.15:31
gundlachjpipes: ah, didn't know it crossed the channel.  ok then.15:31
jpipesgundlach: :)15:31
edaygundlach: jpipes is a true international man :)15:31
* gundlach lived in ireland for a year, didn't get any british culture15:31
gundlacheday: ... of mystery15:31
edaygundlach: some would say15:32
jpipesgundlach: heh, I've always wanted to go to Ireland...never been.15:32
gundlachjpipes: don't bother, it's like britain. ;)15:32
jpipesgundlach: lol15:32
gundlach(in terms of weather == rainy and dreary)15:32
jpipesgundlach: my favorite kind of weather :)15:32
* jpipes was miserable in Austin...15:33
jpipesas dendro-afk and eday can attest to15:33
* jpipes likes his weather like his beer...cold and wet.15:34
* gundlach grew up in the south, on a lake, so the hotter the better15:34
justinsbActually, it's reasonably nice in London at the moment - warmer than San Francisco15:34
gundlachjpipes: and unceasing15:34
jpipesgundlach: well, that too ;)15:34
jpipesjustinsb: yeah?  I'll have to go visit my bro then :)15:34
jpipesif I could put up with a 3.5 yr old for more than a day r two...15:34
justinsbjpipes: It's a good time to escape San Fran - not least because it's about to shut down for burning man.  I took the opportunity to visit my family!15:34
jpipesjustinsb: nice!15:35
justinsbjpipes: Your brother is 3? :-)15:35
jpipesjustinsb: lol15:35
comstudwhere is summer?15:35
comstudit's been damned cold in the easy bay15:35
jpipesjustinsb: actually, my wife and I are going to the 2012 games in London...very excited.15:35
mtaylorjpipes: re: pylint==0.19 ... do we need to upload a new pylint package to the ppa?15:35
jpipesmtaylor: hmm, shouldn't need to..15:35
gundlachcomstud: here15:35
gundlachWeather info for northwest woods, Watkinsville, Georgia (updated on 11:36 AM EDT on August 18, 2010); Temperature: 86.9 F / 30.5 C; Humidity: 76%; Dew Point: 78 F / 26 C; Wind: SE at 19.2 mph / 30.9 km/h; Pressure: 30.09 in / 1018.8 hPa (Rising); Conditions: Clear; Visibility: 7.0 miles / 11.3 kilometers; UV: 7 out of 16; Clouds: Clear (CLR) : -; Yesterday's Maximum: 89 F / 31 C; Yesterday's Minimum: 76 F / 24 C; Sunrise: 6:57 AM15:35
gundlach EDT; Sunset: 8:16 PM15:35
jpipesgundlach: poop.15:36
mtaylorjpipes: ok. didn't _really_ read the traceback... was just scanning15:36
jpipesmtaylor: no prob!15:36
justinsbjpipes: You're a brave man... the London travel situation is the same as ever15:36
* mtaylor hates flying in to London even more than flying inside the US15:36
jpipesjustinsb: oh, I know :) but we can stay with my brother, who is in Birkhamstead...15:36
comstudgundlach, nice minimum15:36
comstudour highs are in the 80s, but lows in the 50s15:36
mtaylorit's the only place where the security situation is _more_ silly than ours15:36
gundlachjpipes: re ppa -- do we currently have 0.19 in the ppa or 0.21.1?15:37
jpipesmtaylor: ++15:37
mtaylorgundlach: I don't think we have pylint in the ppa at all15:37
jpipesyeah, we don't15:37
*** ratasxy has quit IRC15:37
gundlachmtaylor: i don't get ppas, but isn't it supposed to have all of its dependencies?  or does it just list them for a package manager to fetch?15:37
mtaylorgundlach: it's just an apt repo you can add...15:38
mtaylorgundlach: the debian package itself lists the depends... but if one of the depends is something that the normal archive can't supply, then we'll add it to the ppa as well as our packages15:38
jpipesgundlach: they're excellent to add packages that your software has specific dependencies on...15:38
mtaylorgundlach: in a perfect world, the ppa would only have our packages15:38
gundlachmtaylor: oh so you're saying we don't have pylint in the repo as its own project.15:38
gundlachmtaylor: got it, i think15:38
jpipesgundlach: no, it's not in the PPA15:38
*** dendro-afk is now known as dendrobates15:42
*** avsm has joined #openstack15:42
gundlachjpipes: just pushed to lp:~gundlach/nova/pylint-version-update if you don't mind reviewing15:44
edaygundlach: do you think we need really need the rackspace/controllers subdirectory? WHat if it were just api/rackspace/ (with a Controller class in there). controllers in the path seems a bit redundant15:44
jpipesgundlach: will do15:44
gundlacheday: ask cerberus, he set it up and i merged it in.  i'm fine with fewer levels but didn't care enough to change it.15:45
gundlachjpipes: tyvm15:45
*** ratasxy has joined #openstack15:46
gundlacheday: i'd be fine with containing ImageController, ServerController, etc -- Python isn't strict on one class per file15:46
eday_cerberus_: hey! you around?15:48
jpipesgundlach: f=open(pthfile, 'w')15:48
jpipesgundlach: space before and after ;) you fixed the line above it, just not that one :)15:48
gundlachjpipes: fixed in another branch?15:48
gundlachaw, really? smack15:48
jpipess/before and after/before and after =/15:49
*** JimCurry has joined #openstack15:49
jpipesgundlach: hehe, I do it all the time :)15:49
*** lew has joined #openstack15:52
gundlachbzr question:15:53
gundlachwhen i "bzr clone lp:~gundlach/foo foo", why can't I then "bzr push" and have it push to lp:~gundlach/foo ?  Am I doing something wrong that I'm required to respecify the obvious default push target?15:53
jpipesgundlach: bzr branch you mean? :)15:54
gundlachjpipes: oh, is that what i'm doing wrong15:54
gundlachwhat's clone vs branch vs pull? i've tried to read this on the internet and haven't found a clear answer -- probably not looking in the right place15:55
jpipesgundlach: no... you can't push and overwrite a source branch unless you deliberately tell it to (use bzr push --overwrite)15:55
gundlachi'm not trying to overwrite a source branch... i think... i'm trying to merge my changes into a remote repo15:56
jpipesgundlach: bzr branch creates a new branch from another one.  bzr pull will pull in any revisions from the source branch that have occurred since branching.15:56
jpipesgundlach: what's the source branch and what's the target branch? local to Launchpad, or local to local?15:57
gundlachso if i need to make a simple fix to a branch on LP, what are the steps i take?  i don't want to "bzr branch" because i'm just trying to commit a rev to the current branch on LP.15:57
edaygundlach: I think to answer the branch/push Q, bzr uses different vars for the source and dest (branch and push defaults). SO one doesn't set the other. I think :)15:58
jpipesgundlach: you are thinking that the branch on LP that you branched from is the same as your local...they aren't.  they're different.15:58
gundlachjpipes: specifically, i was trying to make the pep8 fix you brought up.  I have lp:~gundlach/nova/install-tools-pep8-fixes, and i had deleted my local branch earlier15:58
gundlachjpipes: ah.15:58
jpipesgundlach: if you deleted your local branch, recreate it (bzr branch lp...)15:58
jpipesgundlach: then bzr push --overwrite15:58
gundlachjpipes: i did clone, which is different in what way?15:59
jpipesgundlach: they are the same (right now)15:59
comstudi think clone is just a synonym for branch in bzr15:59
edayjpipes: well, if he's just adding new revs, he won't need --overwrite, correct? Thats only if you uncommit or something15:59
gundlachi did "bzr clone lp:~gundlach/nova/...pep8-fixes; hackety hack; bzr commit -m "typo"; bzr push"15:59
gundlacheday: erm, i actually did do a bzr revert on my local copy, if that counts as an uncommit15:59
jpipeseday: he's not checking out, he's trying to push to an existing LP location will give an error IIRC16:00
jpipesgundlach: no, that's not an uncommit since it doesn't change anything that bzr cares about16:00
edayjpipes: but if it doesn't change the base revs at all, should be fine16:00
comstudi usually just branch from remote to local16:00
comstudedit local16:00
jpipeseday: ok, you can try it :)16:00
comstudthen push to remote name16:00
jpipescomstud: yup.16:00
comstudno need for --overwrite16:01
*** pvo_ has joined #openstack16:01
*** pvo_ has joined #openstack16:01
*** ChanServ sets mode: +v pvo_16:01
gundlachcomstud: so you have to specify 'bzr push lp:~comstud/path/to/remote/branch' ?16:01
comstudunless a new commit hit remote after you branched16:01
jpipesgundlach: only once16:01
comstudin which case you have to figure out how to pull that in16:01
comstudgundlach- yeah, on the first push only16:01
comstudafter that, it remembers16:01
comstudand you can just do bzr push16:01
gundlachok, fine enough.  it "felt" like it should have inferred that from the fact that i cloned from there, but i can see how if how 'clone'=='branch' then i haven't really specified a push target for my new branch.16:02
jpipesgundlach: it remembers the push location. if you wish to push to another branch  (after already pushing to another) and have bzr remember the new location, use bzr push --remember URL16:02
gundlachjpipes: thanks, i've needed that as well in my few days using bzr16:02
gundlachok, i think i'm straight; back to getting my pep8 branch re-reviewed :)16:02
jpipesgundlach:'s a good one to remember. pun intended. ;)16:02
gundlach(thanks, all)16:03
jpipesgundlach: looks like dendrobates had too quick an approval finger ;)16:04
gundlachyep, after i had already pushed my typo fix, so we're set :)16:04
*** pvo has quit IRC16:04
*** pvo_ is now known as pvo16:04
comstudjay, you wanted me to review ewan's branch?16:05
jpipesmtaylor: I've seen a number of these things pop up...I'm wondering if the test suite is deterministic enough... :(16:05
jpipescomstud: pls do!16:05
comstudok, i'll try to get to it in a bit16:05
jpipescomstud: open to all16:05
comstudconf call right now16:05
comstudi see16:05
gundlachcomstud: fyi i also thought he asked me, specifically, to review it :)  the emails being sent straight to us, rather than to a list, is confusing16:06
mtaylorjpipes: yes. I have too... I think those should be investigated... flaky tests are worse than no tests16:06
jpipesmtaylor: ++16:06
*** pcrews_bbl is now known as pcrews16:06
comstudgundlach, yeah :)16:06
jpipesgundlach, comstud: it's just LP's "personal touch" :P16:07
comstudi felt special.16:07
comstudnow I don't.16:07
* mtaylor goes to file bug16:08
gundlachLaunchpad bug 45823: comstud doesn't feel special.16:09
uvirtbotLaunchpad bug 45823 in xfonts-core "ISO8859-15 fonts are very small (dup-of: 39560)" [Medium,New]
uvirtbotLaunchpad bug 39560 in xfonts-core "The xfonts-*-transcoded packages are missing from Breezy onwards" [Medium,Fix released]
gundlachthanks, uvirtbot16:09
*** jdarcy has quit IRC16:10
*** sophiap has quit IRC16:10
gundlacheday: thx for the approve on pylint-version-update, but i missed the pylintrc in my sed sweep.  is there some more appropriate way than IRC to inform the reviewer that the review and a branch update missed each other like ships in the night?16:11
gundlachps: your review missed my branch update, like a ship in the night.16:11
mtaylorLaunchpad bug 61995416:12
uvirtbotLaunchpad bug 619954 in launchpad "merge request review request of a team should indicate such" [Undecided,New]
mtaylorcomstud: ^^16:12
comstudi don't want to feel special when I shouldn't.16:12
mtaylorI agree16:14
*** sophiap has joined #openstack16:14
mtaylorI had the same reaction to that email this morning...16:14
alekibangoeveryone is so special today...16:15
*** guigui has quit IRC16:16
gundlachmtaylor: while we're on the topic: is it a separate bug that it's hard for me to automatically filter merge proposals into a gmail label?  The only way i can detect them is that the reply-to is set to "mp+<some_number>" -- so i have to make my filter be (reply-to:mp AND  Kinda clunky.16:16
mtaylorgundlach: yup. lemme file that one too16:18
comstudcan't filtering on ^Subject: \[Merge\] ?16:18
comstudi suppose that hits more than you want16:18
gundlachcomstud: i'm in gmail, so i don't think i can do "starts-with" filtering; and yeah, false positives16:18
mtaylorgundlach: does google support filtering on arbitrary headers now?16:18
*** chmouel has joined #openstack16:18
gundlachmtaylor: not aware, sorry16:18
comstudthey come to my gmail too, tho i'm pulling down my gmail into mail.app16:19
* comstud pays attention to this call he's on16:19
gundlachsure he does16:19
alekibangofor me, google is not trustable. they ARE evil.16:20
* gundlach goes ahead and marks his pylint branch as Approved16:21
edaygundlach: I got the pylintrc change in my branch already16:27
gundlacheday: cool; i imagine bzr won't complain when you merge into trunk if you made the same changes i did, right?16:28
edaygundlach: it will reject with a conflict16:29
edaygundlach: I'll merge trunk into mine :)16:29
gundlacheday: if branches A and B both make the same change and try to merge into C, there's a conflict, but if B merges to trunk, then A merges B and then merges to trunk it's OK?16:29
*** maplebed has joined #openstack16:30
edaygundlach: also, when a merge prop changes (like you add a new rev) you can set the status from needs review to work-in-progress and then back to needs review, it will trigger notifications. can also just add a comment asking folks to review new rev  (your Q about how to catch updates)16:30
edaygundlach: correct16:30
gundlacheday: ah, ok, thanks16:30
edaygundlach: so, I can either merge trunk into mine once it lands, or just merge yours directly16:31
edayevery rev has a unique ID on it, so it knows what is/isn't already applied16:31
gundlacheday: i feel like in svn, if two branches are merged into trunk that both do the *same* thing to a given line, the second merge will go without a hitch.  am i wrong?  and if i'm not, why can't bzr do this?16:32
gundlach(sorry that i probably don't grok dvcs enough to understand why my question is dumb :) )16:32
edaygundlach: I don't think I've tried for *exact* changes before, would need to test. I think it would throw a conflict though16:34
*** sparky_ has joined #openstack16:34
edaygundlach: one thing (which I have fixed in my branch) is that pylintrc disable-msg=... needs to be one line (comma separated). multiple lines keep overwriting, so only the last sticks16:35
gundlacheday: ah hah! my mistake.16:36
gundlachlooks like mine's already merged, so shall i just hold off and let your merge fix that?16:36
edaygundlach: yeah, I already have it fixed16:38
*** rajijoom has quit IRC16:40
*** jdmaturen has joined #openstack16:46
*** jdmaturen has joined #openstack16:46
*** zheng_li has quit IRC16:49
JimCurryhappy wednesday openstack!16:51
edayJimCurry: g'morning :)16:52
*** pvo has quit IRC16:53
edayJimCurry: when is it you are coming back to portland?16:53
JimCurrynext week!  working on a babysitter so we can go to dinner.  no candidates so far though….16:54
edayMy wife and I were trying to think of someone we know, couldn't come up with anyone who's around/doesn't already have kids :)16:56
*** jamiew has joined #openstack16:56
*** pvo has joined #openstack16:59
*** ChanServ sets mode: +v pvo16:59
*** jamiew has left #openstack17:01
*** littleidea has joined #openstack17:02
*** zheng_li has joined #openstack17:09
jpipesJimCurry: happy Wednesday to you :)17:11
*** mray has quit IRC17:14
JimCurryI bet Portland and Columbus are better places to be today than 105 degree Austin17:14
*** mrayzenoss has joined #openstack17:21
*** mrayzenoss is now known as mray17:21
*** mray has joined #openstack17:21
*** sirp1 has quit IRC17:22
jerojustinsb: are you around ?17:26
justinsbjero: Hi!17:27
jerohi !17:27
jerojustinsb: I am figuring all the i/o paths in nova, and I was wondering what are the main obstacles with the iscsi support implementation17:28
justinsbjero: Main obstacle is trying to figure out what it should do :-)17:29
justinsbjero: There's a blueprint, but that seems to be a bit of a black hole where ideas go to die17:29
justinsbjero: It was working when it was first written - I expect it now needs some tweaking to merge17:29
justinsbjero: But it's only Linux's OpeniSCSI initiator & target17:30
justinsbjero: And I don't think that's really the most interesting iSCSI use case17:30
justinsbBlueprint here:
justinsbAnd actual 'discussion' here:
jerojustinsb: is it based on the same code as the aoe stuff ?17:32
justinsbjero: Yes.  There's a flag which says whether to use iSCSI or AoE in the volume service17:34
justinsbjero: And then that writes different data which the compute service then uses to determine how to mount the volume17:34
jerojustinsb: so it could potentially be extended to use fiber channel and others17:34
justinsbjero: I'm sure it could.  However, I think we've now established a 'standard' way to do plug-in extension points, so I'd need to check whether this had been done for the volume service yet17:35
justinsbjero: The bulk of the code is here and here
*** rnewson has quit IRC17:36
jerothanks justinsb !17:37
justinsbjero: No problem.  I just checked - it looks like the architecture of the volume service hasn't changed much, so the code should still basically apply.17:38
jerodo you know how volumes are mapped in big cloud infras ? do they use iscsi or FC luns17:38
justinsbIf by 'big cloud infras' you mean EC2, then I don't know.  I think EBS is a home-brew system.17:39
jerorackspacers any hint ?17:39
justinsbI think we have the opportunity to learn from EC2's mistakes here ... EBS doesn't seem to offer compelling advantages over iSCSI.17:40
*** Xenith has joined #openstack17:53
*** kashyapc_ has quit IRC17:59
*** zaitcev has joined #openstack18:03
*** Glace has quit IRC18:07
cwgholt: awake yet?18:09
cwwrt to swift, when adding capacity and rebalancing the ring, i assume care is taken to make sure that no partition has all three physical stores moved?18:12
cwin fact i assume you just do RR across each one so you orphan 1/3 of the data at most?18:12
cw(assuming 3 copies here)18:12
creihtcw: correct18:15
creihtNo more than 1 replica of a partition will be moved18:16
cwone thing that occurs to me, @ 80% utilization ... a rebalance is going to cause a large amount of bi-sectional work amongst the object stores18:18
cwit would be good to have the communications between the stores on a different subnet to the proxy traffic18:18
gholtYeah, that was a consideration here as well, though we also run against aggr port limitations and buying extra ethernet cards, etc. It's a fun balance.18:19
gholtTo get multi-net functionality into swift as it is now should not be hard. Just a bit more dat ain the ring.18:19
cwwell, 1G out of the nodes @ dual 10G agg isn't that bad18:20
cwyeah, jc was telling me it would need some ring work ... i'm just busy at a high level driking coffee and draw lies on a whiteboard18:20
gholtThe only reason it isn't in the code already is that we didn't need it (at this point).18:20
cwring sizing also requires a bit of thought18:20
gholtYeah, my best guess right now is that ring sizing would require a bit of 'read-only' time. But that there might be a way to avoid that.18:21
cwwhen a spindle has a non-unity weight ... does it just take up more space in the ring?18:21
cwgholt: i'm just talking about picking an initial ring size ... 2^19 seems about right planning ahead somewhat18:22
gholtYou mean if a device has a weight of 100 and another has 200? The 200 will get twice the partitions assigned to it.18:22
cwi have 2TiB drives now ... but 3TiB will be out before year end18:22
cwthat's enough of a delta it makes sense to consider it for ring sizing18:23
gholtYeah, that's how that works. One easy way is to just set the weight to the number of Gs.18:23
notmynameone simple thing you could do is assign a weight that is equal to the number of GB on a drive18:23
cwi did some rough math on the whiteboard and found out it would take 150 days to fill up18:23
cwwhich felt very wrong18:24
cwi moved a decimal pt ... real math shows 14 days18:24
gholtI'm horrible at both kinds of math.18:24
notmynametwo weeks to fill up what? your cluster?18:25
cwoh, this is 'astrophysics math' when i whiteboard, you round some up and some down and go for powers of 10 boundaries18:25
cwmostly it works ok18:25
cwnotmyname: potentially yes18:25
cwthat's assuming large objects to get good throughput ... smaller objects would take a lot longer18:26
cwhow do you deal with disk failures?18:28
cwreplace the spindle, remount and let swift deal?18:28
gholtMostly yes.18:29
cwor do you explicitly mess w/ the ring?18:29
cwis the size was the same ... it seems you can [umount], replace disks, mkfs, remount18:29
gholtRing changes really only happen if something can be fixed in a "reasonable" amount of time.18:29
cwand it should JustWork18:29
creihtcw: btw, we've documented some of that in:
cwi know i asked this before ... but do you know the dist of the EA sizes you have in use?18:31
uvirtbotNew bug: #620027 in nova "vgcreate/lvcreate in volume/ fail and go undetected" [Undecided,New]
cwextended attributes18:31
cwcreiht: when you unmount a drive ... what stops swift from writing to the underlying fs?18:31
cwlack of base dir structure?18:32
creihtcw: permissions18:32
gholtcheck_mount calls18:32
creihtcw: and most stuf... yeah what gholt said18:32
cwcheck_mount is done how often?18:32
cwthat sounds expensive ... perms are cheap18:32
creihtcw: Any call that tries to access the dir18:33
gholtWell, os.path.ismount I should say.18:33
gholtIt's pretty cheap too.18:33
cwyeah, until it's all up and under load i'm not so sure i care18:33
cwi'll do some 'perf cpu' testing then18:33
cwi was going to slurp wikipedia into it a test but the data dumps are a pain to deal with18:34
gholtThe check_mount is our function that does a bit more than ismount. swift.constraints.check_mount18:35
*** sandywalsh has quit IRC18:36
gholtLooking at Python's impl of ismount, it's just a couple lstats, so should be quite fast.18:39
*** jc_smith has joined #openstack18:46
*** Xenith has quit IRC18:51
zaitcevWhat's the equivalent of "cvs update -d" or "git pull" in bzr?18:55
gholtafaik bzr pull18:56
*** sandywalsh has joined #openstack18:56
*** thesandywalsh has joined #openstack18:56
zaitcevgholt: Thanks, that worked.18:56
*** thesandywalsh has quit IRC18:57
rbergeronzaitcev :)19:02
zaitcevWhere are my RPMS Robyn19:02
rbergeronyes, i am going to fudcon - I will still let you do all the talking :)19:03
zaitcevI'm installing by cp -a into /usr/local19:03
*** pvo has quit IRC19:03
zaitcevfeeling caveman... with Python!19:03
rbergeronzaitcev: come to #fedora-cloud ;)19:03
rbergeronzaitcev: which rpms are you speaking of specifically19:05
uvirtbotNew bug: #620042 in swift "Add documentation for Debian Packages" [Undecided,New]
zaitcevrbergeron: I thought you were going to package Swift and Nova for Fedora 15 at least.19:06
*** justinsheehy has joined #openstack19:07
rbergeronme personally? nooooooo....19:07
rbergeronswift is almost done - bug 617632 in Package Review "Review Request: openstack-swift - OpenStack Object Storage (swift)" [Medium,Assigned]19:08
rbergeronNova .... is going to be a little stickier, I think, with the dependencies19:08
*** pvo has joined #openstack19:13
*** pvo has joined #openstack19:13
*** ChanServ sets mode: +v pvo19:13
*** pvo has quit IRC19:13
eday_cerberus_: has the prereqs to the api-tests branch, would like to hear your thoughts before I mark to merge19:14
*** pvo has joined #openstack19:18
*** pvo has joined #openstack19:18
*** ChanServ sets mode: +v pvo19:18
*** sirp1 has joined #openstack19:28
_cerberus_eday: looking19:29
*** binaryWarrior has joined #openstack19:31
_cerberus_eday: are list comprehensions considered bad?19:33
_cerberus_I see you refactored one19:33
eday_cerberus_: nope, but long lines are (80 chars or less for pep8)19:36
_cerberus_eday: Ahh yeah, of course19:36
_cerberus_I commented and reviewed as approve19:37
edayI guess I could have just split the comp. into two lines, but I thought this way was a bit more readable19:37
eday_cerberus_: awesome, thanks!19:37
*** avsm has quit IRC19:39
*** justinsb has quit IRC19:48
*** pvo has quit IRC19:53
*** binaryWarrior has quit IRC19:59
letterjmtaylor: there is a merge proposal in the debian branch I would like for you to look at20:00
*** rdw has quit IRC20:02
*** mray has quit IRC20:03
*** mrayzenoss has joined #openstack20:03
*** mrayzenoss is now known as mray20:03
*** mray has joined #openstack20:03
*** thedot has joined #openstack20:16
edaygundlach: duh, obvious way to remove the self.called stuff in tests... just check the result body :)20:25
*** pvo has joined #openstack20:28
*** ChanServ sets mode: +v pvo20:28
gundlacheday: i was wondering if that would suffice for you -- i thought some of them didn't return responses, or maybe you had some bigger plan in mind20:28
gundlachalso whenever i started wondering down that path (sic) I kept getting diverted into trying to figure out a generic class that could do this testing for us, and raise an assert error if something went wrong... and it seemed too complex and i'd throw it out.  probably happened 3 times while reviewing your code :)20:29
edayI'm doing this now:          environ = webob.Request.blank('/test').environ result = Router()(environ, start_response) self.assertEqual(''.join(result), "Router result")20:29
edayerr, bad paste, insert two newlines in there :)20:29
*** sirp1 has quit IRC20:30
gundlachyou could also do self.assertEqual(webob.Request.blank('/test').get_response(Router()).body, "Router result")20:30
gundlachbroken up as needed20:31
vishyi have a lot of the tests passing20:32
*** sirp1 has joined #openstack20:33
vishyin my orm branch20:33
gundlachi made some whitespace changes in tools/ and now volume tests are failing.  are the tests broken?
*** ctennis has quit IRC20:35
edaygundlach: ahh, yeah, I need to read up more on all that is in webob.20:43
*** codejunkie has quit IRC20:44
*** ctennis has joined #openstack20:47
*** londo_ has joined #openstack20:56
*** wreese_ has joined #openstack20:56
*** larsemil_ has joined #openstack20:57
*** zul_ has joined #openstack20:57
*** londo__ has quit IRC20:57
*** larsemil has quit IRC20:57
*** Robi_ has quit IRC20:57
*** thedot has quit IRC20:57
*** jc_smith has quit IRC20:57
*** creiht has quit IRC20:57
*** wreese has quit IRC20:57
*** zul has quit IRC20:57
*** thedot has joined #openstack20:57
*** jc_smith has joined #openstack20:58
*** creiht has joined #openstack20:58
*** ChanServ sets mode: +v creiht20:58
*** Robi_ has joined #openstack20:58
*** allsystemsarego has quit IRC20:59
*** thedot has quit IRC21:03
*** rdw has joined #openstack21:06
*** Podilarius has quit IRC21:11
*** sparky_ has quit IRC21:20
*** ppetraki has quit IRC21:20
*** sirp2 has joined #openstack21:26
*** sirp1 has quit IRC21:27
_cerberus_eday: Where can I find docs for stubout?21:32
edayI just read: pydoc stubout, didn't really look online21:34
*** rnewson has joined #openstack21:34
eday_cerberus_: this has some info too:
eday(it's part of python mox module)21:35
_cerberus_Ahh, thanks21:36
edayoh, wait... thats using stubout through mox, not directly21:36
*** justinsb has joined #openstack21:43
justinsbHi - Launchpad question... I fixed the merge problems on and pushed a new version; how do I request a re-review without starting a new merge request (which would effectively lose all the context)?21:44
*** rnewson has quit IRC21:46
*** e1mer has quit IRC21:47
edayjustinsb: mark as "work in progress" while fixing, and then back to "needs review" once you push21:48
edayjustinsb: can also just leave a comment asking for folks to look at it again21:48
*** JimCurry has left #openstack21:50
edayjpipes: did you see you have text conflicts in your two proposed branches? :)21:50
*** Xenith has joined #openstack21:50
_cerberus_eday: The testing code you have out there so far looks good. Getting pedantic for a second, what you've done so far is effectively mocking, using stubout, since you're ensuring something gets called. I think it's a great start though21:52
mtaylorletterj: looking21:54
mtaylorjustinsb: also, you can click the "resubmit proposal" link, which will create a new merge prop but keep all of the old context21:54
mtaylorjustinsb: sort of depends on how extreme you want to communicate that the new prop is - a resubmit resets any previous votes that were there but keeps the comments. just seting back to "needs review" is a less extreme thing and most of the time what you want21:55
justinsbeday, mtaylor: Thanks21:56
justinsbeday, mtaylor: I tried just toggling the state, but it's a minor merge, so I'll leave a comment as I don't think the toggle did anything (because I'd already pushed)21:56
eday_cerberus_: yeah, I have another rev I'm about to push that cleans up the mocking stuff a bit. definitely want to hear what you think can be improved her21:58
_cerberus_eday: I'll do my best21:58
eday_cerberus_: I may not be taking advantage of mox yet in some way... just not sure how it will help with this21:59
eday_cerberus_: I just pushed another rev to api-tests, pull those down and take a look22:00
_cerberus_eday: there's a lot of philosophical arguing over whether or not you really need to do mocking, since they tend to be very brittle. Any changes in method calls breaks the whole test in what most would argue are useless ways22:00
*** rnewson has joined #openstack22:01
eday_cerberus_: yeah.. I'm not quite sure where I sit for these types of apps yet...22:01
justinsbAnother question... if I were to remove Twisted from the Compute service (replacing with threads) and submitted that, would it be merged?  Should I post this as a Launchpad Blueprint/Bug/Question?22:02
edayjustinsb: if the work is already done, I'd just push it and let folks poke at it :)22:02
mtaylorjustinsb: I'd submit that as a bug or as an email to the mailing list - seems like one of those things that wants discussion22:03
mtaylorjustinsb: OR - do what eday said if it's already done22:03
_0x44eday: Early on we made pretty heavy use of mocks in our current codebase, and that ended up causing us a lot of problems with testing when things would change.22:03
mtaylorletterj: looks good in general. This: rsync (>=3.0.3-2) ... seems a little specific. I believe debian policy is usually to not depend on a specific deb release - so at least make is rsync (>3.0.3) ... unless you _actually_ need debian release -2 for it to work22:04
edayjustinsb, mtaylor: But given the previous discussion on the mailing list re: twisted/threads for these types of workers, it didn't sound like anyone was really against the idea.. certainly would simplify things22:04
justinsbeday, mtaylor: I have a very basic version which is fine for poking at; but I'm sure it won't merge any more...  I'll update it, propose it, and see how long it survives the flames...22:04
eday_0x44: ahh, good to know22:04
mtaylorjustinsb: sounds like a good plan to me!22:04
*** mray has quit IRC22:04
mtaylorletterj: also, you have the override_dh_installdirs but don't override it with anything other than dh_installdirs :)22:05
_0x44eday: Now if someone uses a mock (even inadvertantly) armmer crashes through the wall and crushes them22:05
*** dendrobates is now known as dendro-afk22:06
eday_0x44: do you get kool-aid with the crushing at least?22:06
letterjmtaylor: overriding the dh_installdirs is the only way I could get it to run22:07
_cerberus_eday: only if he tips over22:07
_0x44He makes you drink the kool-aid22:07
_cerberus_And then  you just need a towel22:07
*** e1mer has joined #openstack22:08
mtaylorletterj: fascinating.22:08
*** rbright has joined #openstack22:09
*** gundlach has quit IRC22:10
letterjif you look at a log of the package build (even with *.dirs files in the debian directory)  dh_installdirs is never run.  It is strange.  :-)22:10
letterjmtaylor: I'll make that change and push it up.22:11
mtaylorhrm. well I haven't used .dirs files in a long time since the .install files came into being22:11
mtaylorI'm wondering if there's something we're missing here22:11
justinsbmtaylor, eday: Crud... I can't realistically merge it, because it incorporates some significant other branches (e.g. raw disk images), which have themselves been blocked behind exit code checking :-(  Looks like I'll have to try to get raw disk images through first22:15
eday_cerberus_: do you think the new api-tests changes look less like mocking?22:15
_cerberus_eday: yeah. Like I said, I was just being nit-picky. Verifying that things are called in a specific order *can* be useful but it's usually not.22:16
eday_cerberus_: yeah, seems like all the matters is that the result is correct, not how you got it. the full mock/record/replay thing seems like extra work :)22:19
*** rbright has left #openstack22:20
_cerberus_eday: exactly22:21
*** rnewson has quit IRC22:27
*** spackest has joined #openstack22:41
spackestso for whatever reason my company is rather in love with san and far from in love with locally attached spinning disk22:41
* cw likes local disk very much22:42
spackestwondering if anyone could share some implementation details of their swift setup and how perhaps locally attached spinning disk might actually work for petabytes of data22:42
cwspackest: you have lots of them22:42
cwspackest: if some fail, others have the data22:42
spackestthink we're about to spend (waste :) ) millions of dollars to avoid local disk22:42
spackestany rackspace folks here that can share details?  we could sign ndas if needed22:43
spackestor any other folks with large swift installs?22:43
cwwe're working on a couple here22:43
cwstarting small though22:43
cwi can share concrete details, pics, in the future22:43
spackestthink the thought is that sata fails too often and it is really, really hard to manage replacements22:44
spackestcw: as details come that would be awesome22:44
cwspackest: it's otlerant to failures22:44
cwspackest: what data size do you want to start at?22:44
zaitcevspackest: Have you ever heard of Google? They would have NEVER gotten where they are now if they used SAN.22:45
spackestalmost feels like they think 10% of sata drives fail each day and would require a full team of very technical folks to replace them22:45
spackesteventually 100 PB22:45
cwit's not that bad22:45
cwfrom experience with roughly 30k spindles you get about 1 a day22:45
cwsometimes a little more22:46
cwit's oddly a lot better than i would have expected22:46
*** spackest1 has joined #openstack22:46
spackest1just would like some real data to respond with22:46
*** zheng_li has quit IRC22:47
cw@100PB i can only make up lies22:47
larissacw: Error: "100PB" is not a valid command.22:47
spackest1we're pushing 10 pbs I think but want to grow it22:47
cw  at 100PB i can extrapolate for swift on what we've done so far22:47
zaitcevThere are papers presented at FAST that have real failure data (basically about 3 times worse than claimed MTBF). I don't know why would anyone expect anything different.22:47
cwzaitcev: i want to get some 5900 RPM LP desktop drives and see how those hold up22:48
cwi've been warned off them for now and don't have time to really push it22:48
spackest1just wondering how manageable such a system would be22:48
zaitcevAnd the real trick anyway is not to bother replacing drives. Just leave dead drives there and swap servers when they completely kaput, in a routine sweep.22:48
*** zheng_li has joined #openstack22:49
cwthat sorta sucks22:49
cwyou have parts of the partition that are suboptimal then22:49
*** spackest has quit IRC22:50
spackest1zaitcev: do you have links?22:50
gholtThat actually was a consideration here too. If a drive fails, take it out of the ring. Once a server gets "bad" enough, take it out of the ring, reset everything to new and put it back online.22:51
gholtDepends on how much you trust your drive replacement staff. :)22:51
gholtWe ended up going the replace it sooner just because that makes better use of the resources out there. Sucks to have a beefy machine with half its drives gone.22:52
*** zheng_li has quit IRC22:53
*** zaitcev has quit IRC22:57
*** jdmaturen has left #openstack23:05
cwgholt: our customers have staff on hand in their DCs so the recommendation will be replace it when possible23:05
cwi guess if they drag it out for too long we'll just remove them from the ring23:05
cwi'm a little scared to automate that heavily initially ... in case it blows up :)23:06
gholtSame here. :)23:06
*** miclorb_ has joined #openstack23:07
cwgholt: failure rates increase somewhat as disks age ... but with product growth that ends up being less of an issue hopefully23:08
cwie. most of the stuff you care about is new and shiny23:08
gholtYeah, there seems to be a flurry of just-installed failures as well, even with "proper" burn in. That's make me sad, but it is what it is.23:09
*** pcrews has quit IRC23:09
cwyeah, there are those ... i have someone trying to sell me 1000s of reconditions drives because of that23:10
*** spackest has joined #openstack23:10
cwthey insist they are MORE AWESOME because they have worn down bits or something23:10
*** zaitcev has joined #openstack23:10
*** dendro-afk is now known as dendrobates23:13
*** spackest1 has quit IRC23:13
gholtI have no idea what the lifespan of SSDs would be. Could be interesting.23:14
mtaylorgholt: no one does really23:14
cwthe vendors who have access to the raw bits could probably tell at this stage23:15
cwwhat users see has all the bits corrected ... we can't see how much 'decay' there is inside23:15
cwwith some idea of how that grows over time we could model it to some extent (flash is pretty well understood at this point)23:15
*** sirp2 has quit IRC23:17
*** Kami__ has quit IRC23:20
*** Kami_ has joined #openstack23:21
*** cory_ has quit IRC23:21
*** gasbakid has quit IRC23:24
*** spackest1 has joined #openstack23:32
*** sandywalsh has quit IRC23:33
*** lew has quit IRC23:33
*** littleidea has quit IRC23:34
*** spackest has quit IRC23:35
*** spackest1 has quit IRC23:43
*** littleidea has joined #openstack23:45
edayvishy: any idea what the 'cloud_topic' was used for? It's referenced in the server/unittests, but it looks unused (flag commented out, nothing in nova source sending to it). Am thinking of just removing it as it looks like dead code23:53
*** sirp1 has joined #openstack23:56

Generated by 2.14.0 by Marius Gedminas - find it at!