Thursday, 2017-03-09

« Wednesday, 2017-03-08 Index Friday, 2017-03-10 »
openstackgerritMichael Johnson proposed openstack/octavia master: Look for diskimage-builder not dib-utils when building amps  https://review.openstack.org/44322000:07
openstackgerritMichael Johnson proposed openstack/octavia master: Look for diskimage-builder not dib-utils when building amps  https://review.openstack.org/44322000:07
*** catintheroof has quit IRC00:46
*** ducttape_ has quit IRC00:54
johnsomhttps://review.openstack.org/44333600:59
johnsomThat is so hacky I feel ashamed for posting it....01:00
*** ducttape_ has joined #openstack-lbaas01:13
*** sanfern has quit IRC01:31
*** ducttape_ has quit IRC01:35
*** ducttape_ has joined #openstack-lbaas01:38
*** ducttape_ has quit IRC01:49
rm_workjohnsom: i thought you just needed to add "command -v $DIB_PYTHON || " to the beginning of python_path02:19
rm_workif it's set, great, otherwise do what it did before02:19
rm_workhunting for the package is super hacky and might still not do what you intended02:19
johnsomI tried that first, it failed02:20
rm_workah, hmm02:20
johnsomIt is the version for inside the image02:20
rm_workwell i just commented that. so bleh02:20
rm_workaah, crap02:20
rm_workthis isn't outside02:21
rm_workthere should also be a thing for that, i think?02:21
rm_workhow does anything else do it02:21
johnsomSee the results of the first patch...02:21
johnsomYeah, I spent way too much of my afternoon noodling on this02:22
johnsomThere must be a better way, I just don't know what it is...02:22
johnsomIt is an outside agent that installs the module and the requirements.  We have to figure out which version they used.  Their gates are all py2 on trusty, some of our gates are py3 on xenial02:24
*** gongysh has joined #openstack-lbaas02:28
johnsomMaking risotto so may be off and on for a bit02:28
rm_workjohnsom: so.... uhh02:28
rm_workone thing might be02:29
rm_workjust ignore this temporarily so we can fix the REST of the gate02:29
rm_worksince py3 isn't voting02:29
rm_workIMO02:29
rm_worktake off the depends-on and just ... roll with the patch as-is02:29
johnsomYeah, I could live with that02:29
johnsomYeah, the depends doesn't work anyway02:29
rm_workok i'm going to do that02:29
johnsomOk, ping me and I will run vote on it02:30
johnsomI certainly have reviewed and tested it enough02:30
openstackgerritAdam Harwell proposed openstack/octavia master: Look for diskimage-builder not dib-utils when building amps  https://review.openstack.org/44322002:30
rm_workjust changed the commit message, looks like your changes since i touched it last was net-zero02:31
johnsomCrumb, even the ugly fix didn't work02:37
johnsomIt still picked py3 when the module was installed with py202:37
johnsomOh, I am dumb02:38
johnsom|| true triggers to get around the -e which in turn changes the $?02:38
rm_workerr02:40
rm_worklet me look again02:40
rm_workLOL yes02:41
rm_workif you use true like that02:41
rm_workit will always work02:41
rm_workthat's the point02:41
rm_workyou exactly DO NOT want || true02:41
rm_work:P02:41
rm_worki am dumb too for not catching that02:41
rm_workthough i was more focused on 'why not the simple way'02:41
rm_workif this is INSIDE the amp, it's less hacky02:42
rm_workbecause no chance you will have a random yaml module sitting around02:43
rm_workwhich is what i was thinking about02:43
rm_workso maybe it'll be ok now02:43
johnsomFixed.  Also, I won't shame you if you push that in to get the py2 gates going again02:43
rm_workjohnsom: enjoy your risotto02:43
rm_workyes, +2 plz :P02:43
johnsomI did already02:43
rm_workah you did02:43
rm_work^_^02:43
rm_workyeah i ... uhh02:43
rm_workit's a gatefix02:43
rm_workI can self-merge02:44
rm_workright? ^_^02:44
johnsomYeah, the PTL gives his blessing02:44
rm_workkk02:44
rm_worki'll do that assuming testing is fine02:44
*** ducttape_ has joined #openstack-lbaas02:50
*** armax has joined #openstack-lbaas02:54
*** ducttape_ has quit IRC02:55
*** sanfern has joined #openstack-lbaas02:57
*** amotoki has joined #openstack-lbaas02:57
*** amotoki has quit IRC03:09
*** ducttape_ has joined #openstack-lbaas03:18
*** fnaval has joined #openstack-lbaas03:18
*** ducttape_ has quit IRC03:19
*** ducttape_ has joined #openstack-lbaas03:20
*** ducttape_ has quit IRC03:20
*** ducttape_ has joined #openstack-lbaas03:20
*** bbzhao has quit IRC03:31
*** bzhao has joined #openstack-lbaas03:32
*** amotoki has joined #openstack-lbaas03:32
*** amotoki has quit IRC03:46
*** sanfern has quit IRC04:00
*** amotoki has joined #openstack-lbaas04:02
rm_workok diltram what do you think about https://review.openstack.org/#/c/437200/ ?04:06
rm_workI guess I could do more DB checks or something to make sure none of the objects exist...04:06
*** amotoki has quit IRC04:10
*** KeithMnemonic has quit IRC04:14
*** links has joined #openstack-lbaas04:17
*** amotoki has joined #openstack-lbaas04:24
*** ducttape_ has quit IRC04:28
*** kobis has joined #openstack-lbaas04:35
*** fnaval has quit IRC04:40
*** fnaval has joined #openstack-lbaas04:45
*** kobis has quit IRC04:58
*** sanfern has joined #openstack-lbaas05:00
*** sanfern has quit IRC05:00
*** sanfern has joined #openstack-lbaas05:00
*** sanfern has quit IRC05:01
openstackgerritMerged openstack/octavia master: Look for diskimage-builder not dib-utils when building amps  https://review.openstack.org/44322005:11
*** gongysh has quit IRC05:16
*** ducttape_ has joined #openstack-lbaas05:28
*** ducttape_ has quit IRC05:32
*** fnaval has quit IRC05:40
*** armax has quit IRC05:50
*** kobis has joined #openstack-lbaas05:51
*** kobis has quit IRC05:55
*** gongysh has joined #openstack-lbaas06:01
*** links has quit IRC06:04
*** sanfern has joined #openstack-lbaas06:10
*** links has joined #openstack-lbaas06:16
*** kobis has joined #openstack-lbaas06:19
*** gcheresh_ has joined #openstack-lbaas06:21
*** belharar has joined #openstack-lbaas06:22
*** ducttape_ has joined #openstack-lbaas06:29
*** ducttape_ has quit IRC06:33
*** armax has joined #openstack-lbaas06:39
*** rcernin has joined #openstack-lbaas06:44
*** links has quit IRC06:58
*** amotoki_ has joined #openstack-lbaas07:01
*** amotoki has quit IRC07:03
*** armax has quit IRC07:06
*** ducnc has joined #openstack-lbaas07:10
*** yamamoto has quit IRC07:14
*** links has joined #openstack-lbaas07:15
*** tesseract has joined #openstack-lbaas07:22
*** ducttape_ has joined #openstack-lbaas07:30
*** csomerville has joined #openstack-lbaas07:31
*** cody-somerville has quit IRC07:31
openstackgerritZhaoBo proposed openstack/octavia master: Support vip bandwidth limitation  https://review.openstack.org/44191207:32
*** ducttape_ has quit IRC07:35
*** anilvenkata has joined #openstack-lbaas07:35
*** ducttape_ has joined #openstack-lbaas08:30
*** ducttape_ has quit IRC08:35
*** eezhova has joined #openstack-lbaas08:37
openstackgerritTuan Luong-Anh proposed openstack/octavia master: Indicating the location tests directory in oslo_debug_helper  https://review.openstack.org/44350308:55
*** eezhova has quit IRC09:05
*** krypto has joined #openstack-lbaas09:10
*** krypto has joined #openstack-lbaas09:10
*** Dave____ has joined #openstack-lbaas09:24
*** Dave has quit IRC09:29
*** Dave____ is now known as Dave09:29
*** ducttape_ has joined #openstack-lbaas09:31
*** ducttape_ has quit IRC09:36
*** anilvenkata has quit IRC09:36
*** sanfern has quit IRC09:45
*** eezhova has joined #openstack-lbaas09:55
*** sanfern has joined #openstack-lbaas10:03
*** sanfern has quit IRC10:08
*** krypto has quit IRC10:29
*** gongysh has quit IRC10:44
*** krypto has joined #openstack-lbaas10:58
*** krypt0N has joined #openstack-lbaas11:00
*** krypto has quit IRC11:02
*** belharar has quit IRC11:07
*** krypto has joined #openstack-lbaas11:18
*** krypt0N has quit IRC11:18
*** yamamoto has joined #openstack-lbaas11:29
*** amotoki_ has quit IRC12:07
*** amotoki has joined #openstack-lbaas12:08
*** amotoki_ has joined #openstack-lbaas12:09
*** kobis has quit IRC12:12
*** amotoki has quit IRC12:13
*** amotoki_ has quit IRC12:14
*** yamamoto has quit IRC12:21
*** yamamoto has joined #openstack-lbaas12:21
*** amotoki has joined #openstack-lbaas12:26
*** krypto has quit IRC12:31
*** krypto has joined #openstack-lbaas12:31
*** gongysh has joined #openstack-lbaas12:43
*** krypt0N has joined #openstack-lbaas13:00
*** krypto has quit IRC13:03
*** yamamoto has quit IRC13:07
*** sanfern has joined #openstack-lbaas13:13
*** yamamoto has joined #openstack-lbaas13:13
*** yamamoto has quit IRC13:16
*** yamamoto has joined #openstack-lbaas13:17
*** yamamoto has quit IRC13:21
*** yamamoto has joined #openstack-lbaas13:25
*** chlong_ has joined #openstack-lbaas13:33
*** ducttape_ has joined #openstack-lbaas13:41
*** gongysh has quit IRC14:08
*** amotoki has quit IRC14:08
*** matt-borland has joined #openstack-lbaas14:13
*** ducttape_ has quit IRC14:19
*** krypto has joined #openstack-lbaas14:21
*** krypto has quit IRC14:21
*** krypto has joined #openstack-lbaas14:21
*** krypt0N has quit IRC14:22
*** yamamoto has quit IRC14:23
*** kobis has joined #openstack-lbaas14:23
*** sanfern has quit IRC14:26
*** sanfern has joined #openstack-lbaas14:26
rm_workwtf is with http://logs.openstack.org/36/443336/3/check/gate-diskimage-builder-pep8-ubuntu-xenial/3ee647b/console.html#_2017-03-09_02_41_13_13879514:34
*** links has quit IRC14:35
rm_workstill think these can merge: https://review.openstack.org/43451814:36
rm_workhttps://review.openstack.org/#/c/437200/14:36
rm_workhttps://review.openstack.org/#/c/431179/14:36
rm_workwould love to see those merge... spending a lot of time dealing with tracking them upstream and downstream T_T14:42
*** kobis has quit IRC14:42
*** strigazi has quit IRC14:44
*** strigazi has joined #openstack-lbaas15:04
*** armax has joined #openstack-lbaas15:04
*** ducttape_ has joined #openstack-lbaas15:05
*** chlong_ has quit IRC15:17
*** gcheresh_ has quit IRC15:17
*** yamamoto has joined #openstack-lbaas15:23
*** fnaval has joined #openstack-lbaas15:26
*** yamamoto has quit IRC15:29
*** ducttape_ has quit IRC15:51
*** rcernin has quit IRC15:58
*** KeithMnemonic has joined #openstack-lbaas15:58
*** links has joined #openstack-lbaas16:07
*** ducttape_ has joined #openstack-lbaas16:15
*** kobis has joined #openstack-lbaas16:16
*** eezhova has quit IRC16:18
openstackgerritGerman Eichberger proposed openstack/neutron-lbaas master: Octavia Proxy Plugin  https://review.openstack.org/41853016:24
*** yamamoto has joined #openstack-lbaas16:26
*** yamamoto has quit IRC16:30
*** ducttape_ has quit IRC16:37
diltramrm_work: looks good :)16:46
diltrameven some basic checks will change a lot in this case16:46
diltramwhen we don't have anything real testing it16:46
johnsomActive/Active meeting starting on the hour....16:53
diltramjohnsom: it suppose to be right now16:55
diltramlike in 5 minutes16:55
johnsomYep16:55
diltramresheduled?16:55
diltramrescheduled?16:56
johnsomNo16:56
*** matt-borland has quit IRC16:56
johnsom17:00 UTC...16:57
diltramso it's gonna be in 1 hour or 3 minutes?16:57
johnsomIn 3 minutes16:57
*** matt-borland has joined #openstack-lbaas16:57
diltramok :P16:57
nmagnezijohnsom, tried to join, but looks like webex does not like linux17:05
nmagnezi"Your browser, browser version, or operating system is currently unsupported."17:05
*** yamamoto has joined #openstack-lbaas17:26
*** yamamoto has quit IRC17:32
*** ducttape_ has joined #openstack-lbaas17:37
*** ducttape_ has quit IRC17:38
*** ducttape_ has joined #openstack-lbaas17:38
*** kobis has quit IRC17:39
*** tesseract has quit IRC17:42
*** KeithMnemonic has quit IRC17:47
*** ducttape_ has quit IRC17:49
*** eezhova has joined #openstack-lbaas18:01
*** eezhova has quit IRC18:02
*** links has quit IRC18:04
*** eezhova has joined #openstack-lbaas18:05
*** SumitNaiksatam has joined #openstack-lbaas18:05
johnsomGood stuff for Act/ACt18:10
*** krypto has quit IRC18:15
johnsomWow, all of our talks got accepted per the web site.18:23
johnsomWe are going to be busy18:23
*** eezhova has quit IRC18:27
*** yamamoto has joined #openstack-lbaas18:28
*** gcheresh_ has joined #openstack-lbaas18:31
*** yamamoto has quit IRC18:33
openstackgerritLubosz Kosnik (diltram) proposed openstack/octavia master: Allow to create vip in lb-mgmt-net  https://review.openstack.org/43117918:33
*** armax has quit IRC18:38
xgermanjohnsom - hu?18:43
xgermanall talks?18:43
johnsomAll of them18:44
johnsomHands on included18:44
*** ducttape_ has joined #openstack-lbaas18:50
*** ducttape_ has quit IRC18:55
*** kobis has joined #openstack-lbaas19:13
*** armax has joined #openstack-lbaas19:17
*** SumitNaiksatam has quit IRC19:19
*** yamamoto has joined #openstack-lbaas19:30
*** kobis has quit IRC19:34
*** yamamoto has quit IRC19:35
*** ducttape_ has joined #openstack-lbaas19:51
*** ducttape_ has quit IRC19:56
*** eandersson has joined #openstack-lbaas20:01
eanderssonAny news on lbaas being added to the openstack client?20:01
johnsomYes, it is in plan for Pike.  We created our repository today.  https://governance.openstack.org/tc/reference/projects/octavia.html20:02
johnsompython-octaviaclient20:02
openstackgerritMerged openstack/octavia master: Switch to use exec_command instead of mount/umount  https://review.openstack.org/44272620:14
*** kobis has joined #openstack-lbaas20:21
*** yamamoto has joined #openstack-lbaas20:31
*** yamamoto has quit IRC20:35
*** gcheresh_ has quit IRC20:39
*** chlong_ has joined #openstack-lbaas20:47
openstackgerritMerged openstack/neutron-lbaas master: Switch to use stable data_utils  https://review.openstack.org/44276820:48
openstackgerritMerged openstack/neutron-lbaas master: Switch to use exec_command instead of mount/umount  https://review.openstack.org/44272120:49
*** ducttape_ has joined #openstack-lbaas20:52
*** kobis has quit IRC20:54
*** ducttape_ has quit IRC20:56
*** matt-borland has quit IRC21:02
*** matt-borland has joined #openstack-lbaas21:03
*** yamamoto has joined #openstack-lbaas21:32
*** yamamoto has quit IRC21:38
*** itsuugo has joined #openstack-lbaas21:41
*** itsuugo is now known as aojea_21:47
*** aojea_ is now known as aojea21:50
*** aojea has quit IRC21:51
*** aojea has joined #openstack-lbaas21:52
*** ducttape_ has joined #openstack-lbaas21:52
*** ducttape_ has quit IRC21:57
openstackgerritAdam Harwell proposed openstack/octavia master: Fix cascade delete flow (pool quota requirements missing)  https://review.openstack.org/43720022:03
rm_workweird22:03
rm_workneeds +A again22:03
eanderssonThat's awesome johnsom.22:16
*** ducttape_ has joined #openstack-lbaas22:28
*** kobis has joined #openstack-lbaas22:30
*** kobis has quit IRC22:31
*** matt-borland has quit IRC22:35
rm_workjohnsom: still around?22:38
johnsomYep, in a meeting now though22:39
rm_workk22:39
*** catintheroof has joined #openstack-lbaas22:39
rm_workwtf is going on with https://review.openstack.org/#/c/443336/ -- first it randomly failed on pep8 for no discernable reason, now it failed on the trusty functional check, then rechecked and passed that but failed on centos? >_>22:45
rm_workare the *voting* jobs for DIB unreliable?22:46
ThiagoCMCHey guys, does Octavia works as a container, instead of a VM?22:46
ThiagoCMCMaybe LXD instead of KVM?22:46
rm_workit's in active development to use lxd22:46
ThiagoCMCThat's awesome!22:46
rm_workThiagoCMC: diltram is working on it, hopefully it will work "soon"22:46
ThiagoCMC:-D22:47
johnsomrm_work  The pep8 issue I understood, it has a check for the '-e' I disabled.  So I pushed a fix for that, but the other randoms, yeah, concerning....  Not 100% sure why the trusty gate failed with an issue pulling fedora packages....22:50
johnsomAlso, off my call now so....22:50
rm_worki'm looking at the session stuff currently22:50
rm_worki don't see why the lock session wouldn't pick up the thing that was flushed right before it in the normal session22:50
rm_workbut somehow it doesn't22:50
rm_workor it locks22:51
rm_work:/22:51
johnsomOk.  So, Yeah, that is the question.  It *was* picking it up with the flush for me.  Maybe we need to call an expire (or something similar) on the lock session?  SqlAlchemy has so many nuances with it's session management.22:52
rm_workyeah i was looking22:52
*** eezhova has joined #openstack-lbaas22:52
rm_workchecking how exactly the row is created22:52
johnsomThat query for update is still the first command in the lock_session right?  That could throw it off if not.22:53
rm_worklike, it's already flushing inside the transaction22:53
rm_workso the outer flush should be superfluous22:53
rm_workah the update returns the quota object22:54
rm_workerr22:54
rm_workmaybe that's ok? it's not a locking get()22:54
rm_workhmm22:54
johnsomSo here is how it is supposed to work:22:54
rm_workdo we need a transaction for the lock-session query.withforupdate?22:56
johnsomif there is no quota for the project, use the context session to create an empty project_id row for quotas.  This is an autocommit session, so it should  commit immediately.  I flush the session to make sure.  Then below, using the lock session(non-autocommit) I query for update that row, which should lock that row to do the quota check.22:56
johnsomlock_session is a transaction session22:57
rm_workoh wait, that's exactly what we removed in this patch22:57
rm_workright22:57
rm_workcan we uhh... flush the lock session right before?22:59
*** ducttape_ has quit IRC22:59
johnsomSo, is your use case still "It locks up when there is no entry yet for some project_id in the quota table. So, If you clear out the quota table completely (like from a new DB create or manually), then try to do a single-create as your first call, it will lock in the API.23:01
johnsomIf you do any other thing first to get an entry created, for example do an LB create (and delete it or not, doesn't matter), then you will have a row in `quota` and doing a single-create will proceed."23:01
johnsomIf anything it would be an expire, but that should be the first query on that session, so there shouldn't be any stale data23:03
*** armax has quit IRC23:04
*** ducttape_ has joined #openstack-lbaas23:04
rm_workhmm23:05
*** aojea has quit IRC23:05
rm_workah23:06
rm_workthat update is a subtransaction23:07
rm_workso it commits out of the subtransaction23:07
rm_workbut it's still in a parent transaction23:07
rm_worksee repositories.py line 494+323:07
rm_workjohnsom: ^^23:08
johnsomWhat about it?23:08
rm_workso we have like23:09
rm_worksession / lock_session23:09
johnsomOh, there is another lock_session there23:09
rm_workthen we make session(1) and create the row and flush23:09
rm_workoh?23:09
johnsomline 496, that is probably a problem23:10
rm_workanyway, the new row is now in "session(0)" but THAT is still not flushed?23:10
rm_workerrr what about it?23:10
johnsomOh, we took it out, nevermind.  I'm looking at master in one window and the patch in another.  So, that was already fixed23:11
rm_workoh nm i am wrong, it uses that one23:11
rm_workwe're not in a subtransaction23:11
rm_workfff23:11
rm_worklost again23:12
johnsomOk, so I just loaded up the code the way I think it should be and I want to try to repro23:12
johnsomJust follow those previous comments?23:12
rm_workyeah23:13
rm_workit gets the deadlock on .first()23:14
johnsomYeah, I bet I know what is happening.  It deadlocks on listener right?23:15
rm_workerr23:16
rm_workquotas = lock_session.query(models.Quotas).filter_by(23:16
rm_work                project_id=project_id).with_for_update().first()23:16
rm_workit deadlocks on the .first()23:16
rm_worki split it up23:16
rm_workinto multiple lines to check23:16
johnsomYeah, but doing a single call create, it will pass the LB check, but lock on the listener check23:16
rm_workerr23:17
rm_workit locks on the very first check23:17
johnsomThe second call into check_quota_met23:17
rm_workno23:17
rm_workfirst23:17
rm_workalso:23:17
rm_work2017-03-09 16:17:00.210 19422 DEBUG octavia.db.repositories [req-1aa34152-9085-41cb-b676-8c64fcb46443 - - - - -] q1: <octavia.db.models.Quotas object at 0x4f23a50> check_quota_met /opt/octavia/env/lib/python2.7/site-packages/octavia/db/repositories.py:28723:17
rm_work2017-03-09 16:17:00.210 19422 DEBUG octavia.db.repositories [req-1aa34152-9085-41cb-b676-8c64fcb46443 - - - - -] q2: None check_quota_met /opt/octavia/env/lib/python2.7/site-packages/octavia/db/repositories.py:28823:17
rm_work            q1 = session.query(models.Quotas).filter_by(project_id=project_id).first()23:18
rm_work            q2 = lock_session.query(models.Quotas).filter_by(project_id=project_id).first()23:18
rm_work            LOG.debug("q1: {}".format(q1))23:18
rm_work            LOG.debug("q2: {}".format(q2))23:18
rm_workI put that right before the get for update23:18
rm_workthe lock_session can't see the row23:18
rm_workand it is definitely the FIRST run of check_quotas23:19
rm_work*check_quota_met23:19
rm_workI have a ton of debug logging in there23:19
rm_worki'd know if it ran more than once23:19
johnsomI think we are talking about two different things.  Can you send me the curl you are using to repro?23:20
rm_workcurl -s -X POST -H Content-type:application/json -H "X-Auth-Token:$TOKEN" -d '{"name": "test_singlecreate_lb", "vip": {"network_id": "'$NETWORK'"}, "listeners": [{"name": "test_listener", "protocol": "HTTP", "protocol_port": 80, "default_pool": {"name":"test_pool", "protocol": "HTTP", "lb_algorithm": "ROUND_ROBIN", "members": [{"ip_address": "'$MEMBER1_IP'", "protocol_port": 80}, {"ip_address": "'$MEMBER2_IP'", "protocol_port":23:20
rm_work 80}]}}]}' http://${OCTAVIA_IP}:${OCTAVIA_PORT}/v1/loadbalancers23:20
johnsomThanks23:20
rm_worki can screenshare23:21
*** chlong_ has quit IRC23:21
rm_workthe problem very clearly seems to be that we are not seeing the object be flushed23:23
*** eezhova has quit IRC23:24
rm_workah yeah it's definitely because we session.begin above23:25
rm_workwe flush but it does nothing23:26
rm_workwe need to commit23:26
rm_workumm23:26
johnsomYou removed the lock_session change right?23:26
rm_workwhy are we even doing this session.begin23:26
rm_workyes23:26
rm_workhold up23:26
rm_worki might have it23:27
johnsomThat command does not fail for me23:29
rm_workyep23:29
rm_worki got it23:29
rm_workpushing up the fix23:29
rm_workare you using auth?23:30
johnsomNo23:31
rm_workand did you clear your quotas table?23:31
johnsomYes23:31
rm_work<_<23:31
rm_workwithout auth it doesn't use quotas23:31
rm_workremember?23:31
johnsomRight, darn, noauth.  Ugh23:31
rm_workyeah this is it23:32
rm_workpushing up23:32
rm_workouter session is pointless23:35
*** yamamoto_ has joined #openstack-lbaas23:35
openstackgerritAdam Harwell proposed openstack/octavia master: Fix quota lock on single_create with auth enabled  https://review.openstack.org/43451823:35
rm_workand actually is essentially a second lock-session for these purposes23:35
rm_workisn't necessary'23:35
rm_workhttps://review.openstack.org/#/c/434518/6/octavia/db/repositories.py23:36
rm_worksee removal of session on 495/49323:36
rm_workmaybe this is clearer: https://review.openstack.org/#/c/434518/5..6/octavia/db/repositories.py23:36
johnsomYou were just removing that subtransaction?23:37
rm_worki really wish de-indent was less badly shown23:37
rm_workremoving that session.begin23:37
rm_workbecause the update() does its own23:37
rm_workand having an outer one here23:37
johnsomI agree, it is pointless, but I had to put it in to work around a sqlalchemy bug....23:37
rm_workmeans there's nothing committed23:37
rm_workwhen it's inside check_quota_met it has this session and a lock_session23:38
rm_workwhen it creates the row in session, it isn't committed yet23:38
rm_workbecause it is in a parent transaction23:38
rm_workso lock_session can't see it23:38
rm_workno amount of flushing will help23:38
johnsomAh, that kind of makes sense, but session is autocommit, it should commit instantly...23:39
rm_workthe whole point of session vs lock_session is that one commits immediately23:39
rm_workerr23:39
rm_workautocommit only works if you *don't* session.begin()23:39
johnsomI wish I could remember the sqlalchemy bug that caused me to re-add that.  I know it was a late night, I couldn't figure out why the heck it wasn't working right and when I added that the bug went away.23:40
rm_workmaybe because of the way the lock_session was working then?23:40
rm_workbecause we've changed quite a bit23:40
*** yamamoto_ has quit IRC23:40
rm_worklock_session used to be ... kinda inside this transaction?23:41
rm_workanyway if you test with that it'll work23:41
johnsomYeah, loading it up now23:42
rm_workis diltram or xgerman around?23:45
johnsomnot sure.23:45
rm_worki am, as usual, impatient to get these freaking things merged23:45
rm_workhave one that needs a +A again because jenkins didn't auto-rebase correctly?23:45
rm_worknot sure why23:45
rm_workwait sonar failed why23:46
rm_workugh sec23:46
rm_workforgot to redo tests23:46
rm_workone of them relies on that session i think23:46
rm_workfixing23:46
johnsomYeah, ok, that is working for me23:47
rm_worklol yeah the one that i changed the test23:47
*** ducttape_ has quit IRC23:48
xgermanwassup23:50
* xgerman goes back to playing with kids23:50
rm_workahh ok23:53
rm_workI see the issue23:54
rm_workok so we need a sub-transaction but not THERE23:54
rm_worksec23:54
rm_workoh or we can just use the lock-session I think23:54
rm_workwait, we are23:54
rm_workwtf23:54
rm_workthis test is ... behaving oddly23:57
*** ducttape_ has joined #openstack-lbaas23:58
« Wednesday, 2017-03-08 Index Friday, 2017-03-10 »

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!