19:04:46 <notmyname> #startmeeting swift
19:04:47 <openstack> Meeting started Wed Nov 27 19:04:46 2013 UTC and is due to finish in 60 minutes.  The chair is notmyname. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:04:48 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:04:50 <openstack> The meeting name has been set to 'swift'
19:04:56 <notmyname> may just be a few of us, but at least it will be logged
19:05:00 <notmyname> https://wiki.openstack.org/wiki/Meetings/Swift
19:05:17 <notmyname> chmouel has been working on the swift-bench separation
19:05:24 <portante> nice
19:05:33 <notmyname> I haven't checked on it in about a day. did the last bits land yet/
19:05:35 <torgomatic> does that mean we can remove it from swift now?
19:05:40 * portante slams beer mug on table, roars for another pint
19:05:57 <notmyname> torgomatic: yup, that's the plan
19:06:13 <notmyname> and that's the arbitrary line I'm drawing to have the swift 1.11.0 release
19:06:22 <notmyname> seems like a nice separation point
19:06:57 <portante> any other fixes you want in?
19:07:13 <notmyname> we already got the db locks and the memcache fix
19:07:16 <notmyname> those were key
19:07:28 <portante> agreed
19:07:52 <notmyname> now that we've got early quorum in, the not-as-early quorum seems important
19:08:01 <portante> ;)
19:08:02 <notmyname> I think that merged last night
19:08:14 <notmyname> no-nagle is nice to be in
19:08:15 <portante> late last night. :)
19:08:26 <portante> pun intended
19:08:33 <notmyname> so really, we have a ton of new stuff in 1.11 (including all that above plus ssync)
19:09:16 <notmyname> and I'd also really like to see the swift-bench stuff finished up to psychologically unblock some python-swiftclient patches
19:09:55 <notmyname> portante: torgomatic: acoles: do you have other patches you'd like to see in 1.11?
19:10:02 <notmyname> portante: anything more about diskfiles?
19:10:17 <torgomatic> the capabilities one would be nice
19:10:32 * portante I'd like to fix unit tests, and get the is not None thing in
19:10:33 <notmyname> oh yeah. it would be nice
19:10:55 <notmyname> 10 patch sets, no +2s
19:10:59 <notmyname> https://review.openstack.org/#/c/54857/
19:11:24 <portante> also: https://review.openstack.org/#/c/58188
19:11:28 <notmyname> torgomatic: have you looked at the last patch set since your -1?
19:11:40 <torgomatic> notmyname: no, I've been heads-down in pulling DLO out into middleware
19:11:56 <notmyname> portante: hmm...I looked at that. doesn't look like I added a review
19:12:02 <portante> also: https://review.openstack.org/#/c/56692/
19:12:09 <torgomatic> also, it'd be nice if we got proper stack traces from within threadpool thingies; seems like there should be a way to do that
19:12:26 <notmyname> oh yeah. I +2'd the compeletly broken version ;-)
19:12:32 <torgomatic> something with sys.exc_info() or something, but I don't know how to make that work
19:12:41 <portante> torgomatic: agreed
19:12:46 <acoles> notmyname: i got some more to do on sysmeta, to incorporate what portante started on forcing middleware to start of pipeline, so probably a while longer for that
19:12:54 <notmyname> acoles: ok, thanks
19:13:08 <notmyname> acoles: and you're working with otherjon on that re the account ACls?
19:13:16 <torgomatic> acoles: let me know if you need any help on that; I'm going to need to do some middleware-forcing myself soon
19:13:20 <notmyname> acoles: I'm looking forward to having that
19:13:39 <portante> acoles, torgomatic: also available to help
19:13:45 <acoles> notmyname: yes been in touch with otherjon
19:13:48 <notmyname> great
19:14:24 <acoles> torgomatic: portante: thanks, will shout if I need
19:14:25 <portante> torgomatic: I deferred to acoles on that work, did not want to hold him up, was only working on it because I did not want suggest an idea that acoles would have to go research
19:14:26 <notmyname> portante: I looked at the logging patch briefly earlier this morning. looks like it's just adding the fields to the end of the existing format, right? if so, I'll review in-depth
19:14:40 <portante> notmyname: yes
19:14:52 <notmyname> great
19:15:16 <portante> I am going to address some comments on the coverage improvements for proxy_logging as well
19:16:00 <notmyname> torgomatic: was your comment about tracebacks a volunteer to write a patch? :-)
19:16:15 <torgomatic> notmyname: if I knew where to start, I'd be on it :)
19:16:52 <notmyname> torgomatic: ok :-)
19:17:02 <portante> logging.debug(exc_info=True) is used by python-keystoneslient/middleware/auth_token.py in a few places
19:17:18 <portante> that is what generates all the Tracebacks in the devstack logs that are not attached to an exception
19:17:45 <portante> I'd like to see them stop that, but it is what it is
19:18:23 <torgomatic> yeah, I don't want two related-but-separate tracebacks though; I want one traceback that goes *through* the threadpool plumbing clear down to the error site
19:18:45 <portante> yes, but how they do it might give us some ideas on an approach
19:18:52 <torgomatic> fair enough; I'll take a look
19:18:53 <torgomatic> thank you
19:19:32 <portante> notmyname: I'd really like to see zaitcev's patch land
19:19:47 <notmyname> heh whoops
19:19:50 <notmyname> wrong channel
19:19:51 <portante> it is not a requirement, but right now we have an incomplete "backend" story for implementation
19:19:52 <portante> s
19:20:03 <notmyname> I'll add these three patches (capabilities, timestamps in logging, and cleanup protection) to https://wiki.openstack.org/wiki/Swift/PriorityReviews for 1.11
19:20:07 <notmyname> and I'll make sure there are LP thingies for those that like them
19:20:12 <notmyname> bigger-picture, you all saw my email about general icehouse stuff. the main goal is getting storage policies (ie multi-ring clusters) working. this is what torgomatic peluse and others are working on (some on master and some in the feature/ec branch)
19:20:32 <torgomatic> which is why I'm pulling SLOs and DLOs out into middleware, FWIW
19:20:49 <torgomatic> in case anyone is wondering
19:21:06 * portante whole heartedly supports that ...
19:21:16 <acoles> torgomatic: figured so
19:21:23 <notmyname> torgomatic: gut feeling on needing that for versioned writes too?
19:21:35 <torgomatic> notmyname: a very strong maybe ;)
19:21:44 <notmyname> ok
19:21:53 <torgomatic> I don't know enough about how the versioned writes handle things to decide
19:21:59 <notmyname> torgomatic: did you add anything to launchpad about it, or do I need to?
19:22:05 <notmyname> torgomatic: the code is right there!
19:22:17 <torgomatic> notmyname: I have not touched Launchpad in a while
19:22:39 <notmyname> torgomatic: ok, I'll check if it's covered in what I added for the multi-ring stuff or if soemthing new needs to be added
19:22:54 <notmyname> portante: I too would like to see movement on zaitcev's patch
19:23:39 <portante> how do we make that happen?
19:24:11 <portante> I can beg
19:24:12 <notmyname> I think we'll need a polite nagbot in -swift starting next week
19:24:40 <notmyname> ya, I can start asking for that, and I'll add it to the review list page in the topic
19:25:00 <portante> do you want to make it a gate for 1.11?
19:25:04 * portante hopes
19:25:16 * portante even begs
19:25:17 <notmyname> the dbbroker stuff? I think it would be unlikely
19:25:36 * portante 's hopes are dashed
19:25:38 <notmyname> well, I don't know that it's realistic for getting something late next week
19:26:00 * portante is still dashed against the rocks
19:26:48 <notmyname> portante: you should drink that next pint you asked for earlier. it will make you feel better :-)
19:27:12 <notmyname> anything else to cover about a 1.11 release or current work?
19:27:20 <notmyname> 1.11 first
19:27:28 * portante slams that pint, roars weekly and slovenly for another
19:27:47 <notmyname> anything else about other things?
19:28:00 <torgomatic> if anyone wants to review the DLO functests patch, that'd help me out
19:28:02 * portante looks
19:28:02 <notmyname> acoles: is your federation work still ongoning?
19:28:53 <acoles> notmyname: yes, but i'm focussed on getting th sysmeta stuff done, and i think the DLO SLO moving to middleare will help
19:29:11 <acoles> torgomatic: will review
19:29:11 * portante has a question to gather feedback on
19:29:12 <notmyname> great (/me is happy about that prioritization)
19:29:18 <notmyname> portante: go for it
19:29:28 <portante> regarding middleware environments for logging
19:29:54 <portante> we tried, erroneously, to get the python-keystoneclient auth_token middleware to use the proxy server logger
19:29:55 <notmyname> the log_name/log_facility?
19:30:05 <portante> log_name vs. log_route, I believe
19:30:49 <notmyname> ok
19:30:53 <portante> which made me wonder if we should consider making middleware use the wsgi server's logger in general
19:31:10 <portante> right now every piece of middleware can potentially have its own logger
19:31:30 <portante> but for the admin, that is really all coming from the proxy-server
19:31:51 <portante> each middleware can potentially report entirely different strings in the log files
19:31:51 <notmyname> I can see that it makes sense to default them all to the same place
19:32:02 <torgomatic> yeah, it doesn't really seem necessary to have per-middleware logging configs IMO
19:32:27 <notmyname> as long as you /could/ have it, I'd support changing it to default to use the same thing
19:32:42 <portante> okay, I'll look into that
19:32:46 <notmyname> eg perhaps you want your auth or some debug middleware logging someplace else
19:33:06 <portante> agreed, we can't prevent that
19:33:16 <notmyname> ok, then I support your idea
19:33:23 <portante> we want to just make it so that middleware flows by default with the proxy-server
19:33:26 <notmyname> (it's easy to agree to ideas with no code written yet)
19:33:28 <portante> or object-server
19:33:32 <portante> ;)
19:33:34 <notmyname> ya
19:34:02 <notmyname> anything else?
19:34:14 <portante> one last thing
19:34:16 <portante> sorry
19:34:43 <portante> I have been playing with changes to the proxy server to make the calls to the backends more like an RPC call
19:34:56 <notmyname> rather than http?
19:35:08 <portante> not quiet
19:35:23 <portante> the code would be written to just call a method
19:35:44 <portante> the given implementation of that method might do HTTP, might use local pipe, or something else
19:36:04 <portante> but right now the proxy server code does not have clean line defined for that
19:36:16 <notmyname> if it weren't 5 months away, I'd say that would make a fantastic summit talk
19:36:33 <notmyname> maybe even being 5 months away, it should be ;-)
19:36:45 <portante> with torgomatic pulling out the slo stuff, it seems doable now
19:36:52 <portante> agreed
19:37:06 <portante> hopefully I'll be able to submit a talk for the diskfile stuff and then for the proxy stuff
19:37:06 <acoles> portante: with you re clean line
19:37:28 <portante> if folks have already started that kind of work, I'd be interested
19:37:54 <notmyname> if I hear anything, I'll let you know
19:38:02 <portante> thanks
19:38:09 <portante> that is all I have from here
19:38:17 <notmyname> portante: got a one last last thing?
19:38:26 <acoles> portante: kind of related, eric deliot put  a patch on gerrit to pull a bunch of the http code out of base.py into a library
19:38:39 <portante> oh?
19:38:47 <portante> do you have the id?
19:38:54 <acoles> ...following austin, as an alternative to his refcatoring within base.py
19:39:53 <acoles> portante: give me a minute, i'm on a mobile, will have to search
19:40:01 <portante> no rush, thanks
19:40:03 <portante> acoles
19:40:11 <notmyname> acoles: want to drop in in -swift when you find it?
19:40:24 <notmyname> torgomatic: anything else from you?
19:40:29 <acoles> portante: ok
19:40:31 <torgomatic> nope, I'm good
19:40:59 <notmyname> great. portante: torgomatic: have a good Thanksgiving. acoles: happy thursday
19:41:14 <notmyname> #endmeeting