19:03:27 <torgomatic> #startmeeting Swift
19:03:28 <openstack> Meeting started Wed Jan  8 19:03:27 2014 UTC and is due to finish in 60 minutes.  The chair is torgomatic. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:03:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:03:31 <peluse> excellent
19:03:32 <openstack> The meeting name has been set to 'swift'
19:03:51 <torgomatic> so, the agenda is https://wiki.openstack.org/wiki/Meetings/Swift
19:04:15 <torgomatic> how about we start with sysmeta status? I think the folks here know something about that one
19:04:31 <keving> hi
19:04:33 <torgomatic> #topic sysmeta status
19:05:01 <torgomatic> acoles: anything to report?
19:05:02 <portante> o/
19:05:07 <torgomatic> or anyone, really
19:05:43 <acoles> so the sysmeta patch is close - i hope!
19:05:47 <peluse> must be going well :)
19:06:04 <acoles> i think i have two +1's
19:06:10 <swifterdarrell> acoles: any substantive changes to folks who reviewed it earlier?
19:06:29 <peluse> acoles:  are there other immediate targeted uses beyond the federation thing?
19:06:49 <swifterdarrell> acoles: (I mean I guess I can just look at the patchset--but an overview?  like what happened to gatekeeper functionality)
19:06:56 <acoles> swifterdarrell: no. gatekeeper middleware is back in the patch, same as before. main change is that it now leverages torgomatic;s wsgi manipulation
19:07:10 <swifterdarrell> acoles: cool
19:07:15 <acoles> so should look very familiar to what you reviewd before
19:07:17 <swifterdarrell> peluse: account ACLs will be using it
19:07:23 <peluse> thx
19:07:37 <swifterdarrell> peluse: there's at least one patcset for that which depend on acoles'
19:07:43 <swifterdarrell> *patchset
19:07:51 <peluse> "torgomatic's wsgi manipulation"... is that a drinking game?
19:08:02 <swifterdarrell> peluse: not if you like your liver
19:08:07 <torgomatic> peluse: drink until WSGI makes sense
19:08:26 <peluse> :)
19:08:36 <acoles> if gatekeeper not configured then gatekeeper will be inserted after catch_errors IF catch_errors is first, otherwise gatekeeper is inserted first
19:08:48 <clayg> i think the patch is ready to get in - i'm looking at it - and torgomatic's got a +2 on it
19:08:53 <swifterdarrell> acoles: sounds good to me
19:09:18 * portante wants to look at it, but does not want folks to wait for him to move it forward
19:09:22 <clayg> if anyone else really wants to get eyes on it they might stick a -1 on there so I don't approve it when/if I +2
19:09:23 <peluse> good stuff acoles
19:09:51 <clayg> swifterdarrell: I was still hoping otherjon might +1 it since his patch depends on it
19:10:17 <acoles> also, object sysmeta support is not included as discussed in channel, that is pending figuring out semantics of POSTing changes to symeta on objects. just saying,
19:10:23 <swifterdarrell> clayg: ya, but I think he'll just adjust/rebase to whatever lands--esp if it's very similar to hwo it used to be
19:10:41 <torgomatic> yeah, doing persistent metadata on objects is kind of hard
19:10:41 <clayg> swifterdarrell: it's even BETTER
19:10:49 <swifterdarrell> :)
19:11:36 <clayg> torgomatic: i don't see how making the sysmeta of an object tied to like of that timestamp (instead of the life of the resources) negates the usefulness of restricted sys meta on objects
19:11:41 <swifterdarrell> torgomatic: acoles: cool--i'm perfectly happy punting on obj sysmeta since there's clear value in acct/container getting delivered sooner/with less effort
19:11:50 <clayg> but... we can do that, and consolidate after_fn, later
19:12:20 <clayg> i haven't found anything on the patch yet worth blocking it any longer
19:13:06 <acoles> clayg: one issue is different bits of middleware competing to update subsets of sysmeta
19:14:34 <swifterdarrell> torgomatic: acoles: clayg: k;  is that sufficient for the agenda item (status update)?
19:14:39 <torgomatic> good enough for me
19:14:41 <clayg> next topic!
19:14:53 <torgomatic> #topic python-swiftclient status
19:15:04 <torgomatic> who knows about swiftclient's status?
19:15:25 <swifterdarrell> torgomatic: there's some require-SSL patch that's really important and languishing
19:15:43 <torgomatic> swifterdarrell: oh yeah; didn't that patch just flat-out break stuff for some folks?
19:15:45 <swifterdarrell> torgomatic: I had problems w/it w/some self-signed cert I'd created a while back; so I didn't like it
19:16:16 <swifterdarrell> But someone said I was crazy, and I think they might have been right, so I just went and did other things for a while
19:16:24 <swifterdarrell> that's all I know
19:16:48 <torgomatic> beyond that, it seems like there's a lot of patches for py3 support in the queue
19:16:49 <clayg> swifterdarrell: fwiw I think the patch could easily add the feature without breaking backwards compat
19:17:23 <clayg> swifterdarrell: have an ENVVAR and add the feature - then change the default whenver we *want* to bump the rev
19:17:45 <swifterdarrell> clayg: the allow-insecure-ssl thing?
19:17:59 <clayg> i told this to notmyname and he said "na" so I stayed out of it
19:18:28 <clayg> swifterdarrell: yeah
19:18:58 * swifterdarrell shrugs; I just want --insecure to, well, allow all valid SSL certs (where in my crazy expereience, it didn't)
19:19:15 <notmyname> ....and my alrm didn't go off.. soory
19:19:21 <torgomatic> he lives!
19:19:24 <swifterdarrell> I like the default-verify behavior change
19:19:35 <swifterdarrell> that part is probably quite solid
19:19:43 <openstack> notmyname: Error: Can't start another meeting, one is in progress.  Use #endmeeting first.
19:19:45 <torgomatic> swifterdarrell: I'm with you on that one
19:19:49 <clayg> see
19:19:52 * clayg stays out of it
19:19:54 <notmyname> ah, cool
19:20:21 <swifterdarrell> notmyname: torgomatic started; we finished sysmeta status update and are on python-swiftclient
19:20:39 <notmyname> yup, just glanced over the backlog
19:20:51 <torgomatic> linear agenda ordering is for chumps ;)
19:21:26 <notmyname> I'm glad I actually looked at the agenda then
19:21:32 <notmyname> or put stuff on it
19:21:40 <torgomatic> related to swiftclient, how's the progress on removing swiftclient as a Swift dependency coming?
19:21:42 <swifterdarrell> So are there any action items here for python-swiftclient?  I guess I could look at teh ssl thing again, but the most recent prob was one someone else (at HP?) had, so i dunno what I could add
19:21:48 * portante loves watching the team mess with notmyname's head in Perth
19:22:18 <notmyname> chmouel was working on python-swiftclient things
19:22:23 <notmyname> chmouel: around?
19:22:32 <torgomatic> I saw a commit merge: 150f338 Remove swiftclient dep on direct_client
19:22:41 <creiht> his last patch got through
19:22:54 <notmyname> ok, great
19:23:37 <notmyname> I'll look into where swiftclient is still used and go from there
19:23:45 <torgomatic> looks like container-sync and dispersion-report are what's left
19:23:55 <torgomatic> (and functional tests, but swiftclient is fine in there)
19:24:06 <notmyname> ok
19:24:16 <cschwede_> i could look into dispersion-report if that helps
19:24:42 <notmyname> cschwede_: thanks. that'd be great
19:25:37 <torgomatic> anything else on swiftclient?
19:25:51 <notmyname> not from me
19:26:33 <torgomatic> #topic log #openstack-swift or not
19:26:41 <notmyname> ok, this one is mine
19:27:03 <notmyname> the question is as the topic says. should we publicly log #openstack-swift
19:27:11 <portante> on this topic, I have had my comments in #openstack-swift picked up and noticed by the powers at be today already
19:27:16 <clayg> notmyname: NOW can we have a vote!?
19:27:20 <notmyname> heh
19:27:22 <creiht> lol
19:27:35 <creiht> notmyname: I don't see why not
19:27:49 <notmyname> portante: so this was exactly my concern. I know that's happened to you, so do we want to have the logs existing and searchable?
19:27:51 <peluse> I vote sure
19:28:23 <torgomatic> hang on hang on
19:28:24 <portante> I don't personally mind that, as I consider this semi-permanent anyways
19:28:26 <swifterdarrell> The channel's already public, so maybe it makes it easier to find, but not any more public
19:28:47 <torgomatic> #startvote Should #openstack-swift have public logs? yes, no, abstain
19:28:48 <openstack> Begin voting on: Should #openstack-swift have public logs? Valid vote options are yes, no, abstain.
19:28:49 <openstack> Vote using '#vote OPTION'. Only your last vote counts.
19:28:54 <notmyname> for others, what's happened (without too many details) is that portante has had others withing Red Hat contact him about something he said in channel
19:28:57 <torgomatic> AUTOMATION! :)
19:29:04 <peluse> #vote yes
19:29:08 <swifterdarrell> #vote yes
19:29:12 <creiht> #vote yes
19:29:14 <torgomatic> #vote abstain
19:29:18 <clayg> #vote yes
19:29:18 <portante> #vote abstain
19:29:19 <keving> #vote yes
19:29:25 <acoles> #vote yes
19:29:33 <cschwede_> #vote abstain
19:29:36 <fbo> #vote yes
19:30:20 <portante> are other openstack channels logged?
19:30:24 <torgomatic> #endvote
19:30:24 <openstack> Voted on "Should #openstack-swift have public logs?" Results are
19:30:25 <openstack> yes (7): peluse, acoles, keving, swifterdarrell, creiht, clayg, fbo
19:30:26 <openstack> abstain (3): cschwede_, portante, torgomatic
19:30:31 * notmyname is internally debating
19:30:31 <creiht> portante: yes
19:30:38 <notmyname> heh, too late
19:30:39 * clayg guesses it was probably like "he portante saw you were talking and generally being a baddass - keep it up"
19:30:40 <torgomatic> notmyname: oop, sorry, jumped the gun there
19:30:48 <creiht> clayg: lol
19:31:03 <portante> clayg: unfortunately, no.
19:31:09 <notmyname> torgomatic: no worries
19:31:10 <creiht> http://eavesdrop.openstack.org/irclogs/
19:32:00 <creiht> it would be better to be transparent
19:32:05 <cschwede_> portante: i think all (major) except ... swift :)
19:32:15 <portante> ah
19:32:26 <notmyname> BTW, if anyone (else) ever has someone approach them trying to limit what's said in any way in openstack IRC channels, I'd love to argue with them about it.
19:32:43 <notmyname> *them == the person trying to limit what's said
19:32:58 <torgomatic> really, some folks already have logs
19:33:23 <notmyname> I'm fine with logging. I just didnt' want to turn it on without mentioning it in a meeting first (and this is our first meeting since it last came up)
19:33:27 <torgomatic> I mean, I know my machine has them saved somewhere
19:33:37 <notmyname> ok, moving on :-)
19:33:39 <torgomatic> so, anyone in the channel, plus anyone who works for the NSA
19:33:51 <peluse> well, that's given
19:33:55 <portante> sssh maybe they are sleeping ...
19:34:39 <torgomatic> #topic Swift 1.12.0 release
19:34:50 <notmyname> we did sysmeta already?
19:35:08 <portante> yes, acoles is doing a great job
19:35:19 <notmyname> yay. good job acoles!
19:35:21 <notmyname> :-)
19:35:27 <notmyname> ok, swift 1.12.0 release
19:35:33 <notmyname> here's what I'm thinking:
19:35:35 <acoles> i got a bunch of help
19:36:06 <notmyname> when the sysmeta stuff lands, I'd like to seriously look into a 1.12 release
19:36:18 <notmyname> so, in the near term
19:36:55 <notmyname> the sysmeta functionality + (hopefully) swiftclient extraction + (maybe) account ACLs would be very nice for a release (+ the other stuff that's landed)
19:36:57 <swifterdarrell> notmyname: can taht include account ACLs plz?
19:37:07 <notmyname> yes, I hope so
19:38:07 <notmyname> I haven't had a chance this week to sit down and look at the specific details of what will be in it, but I wanted to at least have a near-term marker for doing it
19:38:12 <clayg> swifterdarrell: notmyname: I don't think account acl's are going to just fall in once sysmeta lands?  I'm sure much fewer people have been looking at that because of the patch chain dependency
19:38:27 <swifterdarrell> clayg: I hope so :)
19:38:34 <swifterdarrell> clayg: as does otherjon, no doubt
19:39:18 <swifterdarrell> clayg: wait, I misread what you typed
19:39:34 <notmyname> clayg: that's true, but the broad parts of account acls should be pretty good, I think. he's done a lot of work to ensure that the overall design is in the right direction
19:39:45 <swifterdarrell> clayg: I'm hoping the ACL biz is straightforward w/a solid foundation of an already-merged sysmeta
19:40:30 <swifterdarrell> clayg: because, if not, assuming the rough edges weren't related to actually persisting sysmeta, we should have already banged out any rough edges on the account ACL stuff concurrently w/the sysmeta patch
19:41:21 <clayg> swifterdarrell: I just don't think anyone else has looked at it, I've mostly compromisied to the approach as disccused (but that wasn't what was in the patch last I looked)
19:41:50 <swifterdarrell> clayg: k... we'll see how it shakes out :)
19:41:51 <cschwede_> swifterdarrell: i had a look at it and have a question/idea about it, but we can talk after the meeting on #openstack-swift
19:41:54 <clayg> eitherway, all of the same stuff (container vs. account format, json in headers, write vs read-write) is probably going to come up again in review
19:42:02 * clayg is assuming someone else *will* look at it
19:42:21 <clayg> oh yeah!  i forgot cschwede_ was looking at it!
19:42:24 * clayg hugs cschwede_
19:42:25 <swifterdarrell> cschwede_: sounds good; otherjon is the man to talk to, btw
19:43:01 <swifterdarrell> notmyname: torgomatic: sorry for the de-rail; back to 1.12.0?
19:43:19 <notmyname> any other questions on 1.12?
19:43:38 <clayg> notmyname: what's the cutoff for account acl's?
19:43:55 <clayg> notmyname: i'm pretty sure sysmeta is close - i don't really understand what needs to happen with swiftclient
19:44:12 <clayg> notmyname: are you just waiting for a list of features to land then then it goes?
19:44:32 <notmyname> there isn't a cutoff yet. but in-general, let's say end of next week would be a good time to have things for 1.12 landed
19:45:08 <swifterdarrell> notmyname: ouch; that's a tight window for the acct acl IMO
19:45:41 <clayg> lol, I thought I wsa playing the role of pessimist todate?
19:45:43 <notmyname> so let's make sure to review it :-)
19:46:54 <notmyname> that's a goal, bit yet a requirement (the target timeframe fro 1.12)
19:46:58 <swifterdarrell> clayg: it's a team effort
19:47:11 <swifterdarrell> notmyname: k
19:47:26 <creiht> clayg: within a group full of pessimists, yes even you can look like an optimist at times :)
19:47:44 <notmyname> heh
19:47:57 <clayg> swifterdarrell: notmyname: otherjon: ummm... does anyone have a link to the account acl patch?
19:48:29 <swifterdarrell> clayg: it's probably on page 4 in gerrit?
19:48:35 <swifterdarrell> (haha)
19:48:47 <cschwede_> https://review.openstack.org/#/c/63227/
19:48:54 <notmyname> thanks
19:49:11 <clayg> wow that commit msg is epic
19:49:17 <swifterdarrell> https://review.openstack.org/#/c/63227/
19:49:48 <notmyname> so please go take a look a that
19:49:52 <swifterdarrell> #link https://review.openstack.org/#/c/63227/
19:50:22 <portante> is that on the priority review list?
19:50:37 <notmyname> any other 1.12 questions? I want to talk about the gate briefly here at the end of the meeting
19:50:37 <notmyname> torgomatic: topic open discussion
19:50:44 <torgomatic> #topic open discussion
19:50:51 <notmyname> portante: maybe? I hope
19:50:59 * portante is being lazy
19:51:04 * notmyname hasn't looked at the priority review list this week in AUS yet
19:51:10 <notmyname> torgomatic: thanks
19:51:14 <torgomatic> gate queue is 93 deep; 8 more and it breaks the graph ;)
19:51:22 <notmyname> so, as everyone knows, the gate is really backed up
19:51:24 <portante> it hit 101 today
19:51:26 <notmyname> http://not.mn/gate_status.html
19:51:29 <portante> that I saw
19:51:34 <notmyname> wow
19:51:44 <portante> but it needs to say over 100 for a while in order to make it into the graph, though, right?
19:51:45 <cschwede_> looks like the queue is stopped
19:52:05 <notmyname> ya, it's an average of the time-bucket (12 hours currently)
19:52:36 <notmyname> jeblair and clarkb are here at LCA and I know they've been working on getting jenkins performing better
19:52:57 <cschwede_> no started zuul jobs on http://status.openstack.org/zuul/ for the last hours
19:53:07 <notmyname> they're working hard on their side of it (jenkins just got too overloaded--something like a load average of 350)
19:53:35 <notmyname> but that's a small part of "there are >90 things in the gate and gate resets keep happening frequently"
19:54:10 <portante> i asked the rh openstack storage team if they are feeling the pain
19:54:12 <notmyname> so, I'd like to ask for continued patience with the -infra team for their part. they are actively working on getting the queues flowing
19:54:17 <portante> and they emphatically said yes
19:54:18 <notmyname> portante: and?
19:54:24 <notmyname> good to know
19:54:30 <portante> but nobody volunteered to take any action
19:54:37 <notmyname> the fact that gate jobs fail frequently is a separate concern
19:54:43 <portante> folks seem at a loss as to what to do
19:54:47 <notmyname> so I wanted to make that clear
19:55:03 <notmyname> portante: unfortunately, a lot of it is "try harder to not write bugs"
19:55:21 <creiht> time to fork and land large pachsets each night? :)
19:55:31 <notmyname> but the topic came up again in this weeks openstack meeting
19:55:39 <portante> but they said they are spending a lot of time triaging failures in obscure code somewhere in openstack instead of regular work
19:56:06 <notmyname> I'm concerned about the additional dev time cost like that
19:56:28 <creiht> I'm concerned that they still think they can make this all work :(
19:56:40 <notmyname> we see it too. we babysit patches and it take a lot of time instead of reviewing and writing swift code
19:56:50 <portante> it is too much, materially costing companies money to participate that they did not on and cannot plan for
19:57:03 <portante> creiht: agreed
19:57:09 <notmyname> 4 minutes....typing faster
19:57:30 <notmyname> creiht: so while I agree to some extent, it's not "us" vs "them"
19:57:37 <creiht> oh I know
19:57:56 <notmyname> we, as a community, need to figure out how to unclog the pipes
19:58:19 <creiht> lol
19:58:26 <portante> the voice of reason from Perth
19:58:31 <notmyname> 2 thigns we've talked about are automating the recheck comments and monitoring the gate queue resets
19:58:44 <notmyname> and we can actually do those things
19:58:46 <creiht> that's the problem though... there is no community there, it is only "them" who knows how to do it right
19:58:47 <portante> torgomatic does that now for us. :)
19:59:25 <notmyname> creiht: I think there is a lot of pain felt by a lot of people. and many people are working on it
19:59:51 <notmyname> I'd like to see some changes in how integration tests work, but that didn't go so well when we presented at the team meeting.
19:59:51 <torgomatic> time's just about up, fwiw, but we can keep talking in #openstack-swift
20:00:06 <torgomatic> ...and there it goes
20:00:06 <creiht> notmyname: yes, as there have been for a long time.  How long will we continue to bang our heads against the wall until we realize that _perhaps_ a new approach may have to be considered
20:00:11 <creiht> heh
20:00:11 <torgomatic> #endmeeting