Friday, 2014-10-31

*** eglynn has quit IRC00:01
*** zul has quit IRC00:10
*** amalagon has quit IRC00:42
*** ddieterly has joined #openstack-ceilometer00:57
*** _cjones_ has quit IRC01:04
*** amalagon has joined #openstack-ceilometer01:05
*** ddieterl_ has joined #openstack-ceilometer01:11
*** ddieterly has quit IRC01:13
*** ryanpetrello has joined #openstack-ceilometer01:14
*** boris-42 has quit IRC01:23
*** amalagon has quit IRC01:29
*** ryanpetrello has quit IRC01:35
*** amalagon has joined #openstack-ceilometer01:37
*** alexpilotti has quit IRC01:38
*** amalagon has quit IRC01:42
*** ryanpetrello has joined #openstack-ceilometer01:58
*** yasu_ has joined #openstack-ceilometer02:02
*** ryanpetrello has quit IRC02:02
*** nosnos has joined #openstack-ceilometer02:07
*** zul has joined #openstack-ceilometer02:29
*** harlowja is now known as harlowja_away02:31
*** Guest51339 is now known as mgagne02:38
*** mgagne has joined #openstack-ceilometer02:38
*** promulo_ has quit IRC02:45
*** gordc has quit IRC03:12
*** ccrouch has left #openstack-ceilometer03:15
*** nosnos has quit IRC03:27
*** nosnos has joined #openstack-ceilometer03:28
*** nosnos has quit IRC03:32
*** nosnos has joined #openstack-ceilometer04:01
*** nicknach has left #openstack-ceilometer04:16
*** asalkeld has quit IRC04:19
*** sbfox has joined #openstack-ceilometer04:32
*** yasu_ has quit IRC05:14
*** yasu_ has joined #openstack-ceilometer05:32
*** asalkeld has joined #openstack-ceilometer06:11
*** _cjones_ has joined #openstack-ceilometer06:33
*** _cjones_ has quit IRC06:33
*** _cjones_ has joined #openstack-ceilometer06:34
*** k4n0 has joined #openstack-ceilometer06:34
*** nosnos has quit IRC06:34
*** asalkeld has quit IRC06:35
*** ajayaa has joined #openstack-ceilometer06:55
*** _cjones_ has quit IRC07:06
*** nosnos has joined #openstack-ceilometer07:09
ajayaajd__, Hi. I was going through the notification handler for cinder in ceilometer. Here I see that When a volume is created two samples are stored for volume.size meter and similarly when a volume is deleted two samples are created for volume.size. Are we not storing three samples extra for volume.size meter?07:11
ajayaaeglynn-office ^^07:12
*** nosnos has quit IRC07:40
*** nosnos has joined #openstack-ceilometer07:43
*** eglynn has joined #openstack-ceilometer07:56
*** sbfox has quit IRC08:18
*** ildikov has joined #openstack-ceilometer08:20
*** ifarkas has joined #openstack-ceilometer08:33
*** nellysmitt has joined #openstack-ceilometer08:39
*** admin0 has joined #openstack-ceilometer08:46
*** janisg has joined #openstack-ceilometer08:48
*** Yanyanhu has joined #openstack-ceilometer08:51
*** safchain has joined #openstack-ceilometer08:57
*** eglynn has quit IRC08:59
*** isviridov_away is now known as isviridov08:59
*** _nadya_ has joined #openstack-ceilometer09:32
*** boris-42 has joined #openstack-ceilometer09:46
*** cdent has joined #openstack-ceilometer09:50
eglynn-officeajayaa: good morning09:50
ajayaaeglynn-office, good morning! It's afternoon in India btw. ;)09:51
eglynn-officeajayaa: so, we've two volume notification handlers, one of which emits "volume" sample (i.e. the thing exists)09:52
eglynn-officeajayaa: ... and the other emits a "volume.size" sample (i.e. so many GB in block storage is accounted for by this volume)09:52
eglynn-officeajayaa: the reason you're seeing 2 samples in quick sucession for each meter09:52
eglynn-officeajayaa: ... is I suspect because the volume notifications occur in pairs09:53
eglynn-officeajayaa: ... i.e. volume.delete.start forllowed by volume.delete.end to bracket off the deletion workflow09:53
eglynn-officeajayaa: I suspect we should only react to volume.delete.end in this case09:54
ajayaaeglynn-office, Yes. That's what I was thinking also. The volume.size meter should only catch one of volume.create.start, volume.create.end, volume.delete.start, volume.delete.end.09:55
ajayaaThe most appropriate I think is volume.create.end09:55
eglynn-officeajayaa: yes, volume.create.end amd volume.delete.start09:56
eglynn-officeajayaa: *and09:56
eglynn-officehmmm, IIRC we used to have a similar pattern with the comoute event types09:56
*** IvanBerezovskiy has joined #openstack-ceilometer09:57
eglynn-office... i.e. just consuming compute.instance.create.end as opposed to compute.instance.create.{start|end}09:57
* eglynn-office digs in git fossil record09:57
ajayaaeglynn-office, okay. Can I file a bug for this?09:57
eglynn-officeajayaa: sure, please do09:57
ajayaaeglynn-office, Thank you.09:58
eglynn-officeajayaa: yes, that was the case initially for the compute.instance.{create|delete} events ... https://github.com/openstack/ceilometer/commit/d615fb8710:00
eglynn-officehmmm10:00
* eglynn-office dig some more10:01
*** openstackgerrit has joined #openstack-ceilometer10:01
ajayaaeglynn-office, I am not convinced that volume.size meter should react to volume.end.start. The meter "volume" should react to both volume.create.end and volume.delete.start10:03
*** _cjones_ has joined #openstack-ceilometer10:03
ajayaaThe volume.size is only concerned with size of volume of created.10:04
*** ildikov has quit IRC10:06
openstackgerritAjaya Agrawal proposed a change to openstack/ceilometer-specs: Add spec for MagnetoDB metering support  https://review.openstack.org/12633510:06
eglynn-officeajayaa: volume.end.start?10:06
liushengHi eglynn-office, ajayaa , for samples from notification, gordon chung has filed a bug https://bugs.launchpad.net/ceilometer/+bug/138487410:06
eglynn-officeajayaa: we should react to volume.create.end I think10:06
liushengthat try to drop the meters that are really events10:06
ajayaaeglynn-office, yes!10:07
eglynn-officeajayaa: that much block storage is now available for use once volume.create.end is emitted10:07
*** _cjones_ has quit IRC10:08
ajayaas/volume.end.start/volume.delete.start10:08
ajayaaeglynn-office, +110:08
eglynn-officeajayaa: agreed ... I'm digging  thru history to figure out (un-forget) why we changed this10:09
ajayaaeglynn-office, liusheng, Can you please point me to a place where I can learn more about events.10:10
eglynn-officehmmm, here's the change I was looking for10:12
eglynn-officehttps://review.openstack.org/#/c/38485/7/ceilometer/volume/notifications.py10:12
shardyHey Ceilometer folks10:12
eglynn-officeshardy: hey!10:12
shardyFYI I've raised this spec about proxying metric samples via heat:10:13
shardyhttps://review.openstack.org/#/c/132198/10:13
shardyComments re the sanity or otherwise of the approach would be great! :)10:13
liushengajayaa: for instance create/delete events, i.e. notification-agent will emit samples "instance", but this meter actually is events, we have polling agent to gather instance samples10:13
shardyhey eglynn-office!10:13
shardyNot urgent, I know everyone is super busy pre-summit, just wanted to draw your attention  to it :)10:14
eglynn-officeshardy: interesting, will definitely read before summit ... we can discuss f2f next week if needed10:14
shardyeglynn-office: Yup, that would be great, wanted to get the idea out there ahead of time10:14
shardyeglynn-office: thanks!10:15
eglynn-officeshardy: makes sense :)10:15
eglynn-officejd__: can you remember why https://review.openstack.org/38485 changed from consuming volume.start.end & volume.delete.start, to volume.create.* & volume.delete.*?10:16
eglynn-officejd__: seems were doubling up the number of samples triggered by a create and a delete10:16
eglynn-officejd__: (similarly for compute.instance.create.start & compute.instance.delete.end)10:17
shardyeglynn-office: Thanks for the feedback10:18
shardyeglynn-office: the only risk I see is all the agents become a giant Heat DoS engine, so I guess we'll have to selectively rate limit the signals somehow10:19
shardyeglynn-office: does ceilometer rate-limit sample creation at the API level?10:19
*** Yanyanhu has quit IRC10:20
eglynn-officeshardy: yep, that's a good point ... distributed rate limiting is not something we've been strong on traditionally in OpenStack10:20
eglynn-officeshardy: nope, ceilometer-api does no rate limiting as things stand10:20
eglynn-officeshardy: this has been raised as a customer issue10:20
shardyeglynn-office: Yeah, it should be a relatively inexpensive thing for heat to handle, but it's still a concern10:20
*** ildikov has joined #openstack-ceilometer10:21
eglynn-officeshardy: how does heat do it, if the heat-api/engine layers are horizontally scaled out?10:21
shardyThe problem for heat is we might be willing to drop a few metric signals on the floor, but we must never ever drop completion signals etc10:22
shardyeglynn-office: currently we don't really do rate limiting either tbh10:22
shardyour main problem is when other services do, and we need to retry things10:22
*** underyx|off is now known as underyx10:22
eglynn-officeshardy: I seem to remember vek up spinning a turnstile project to do distributed per-tenant rate limiting across scaled out nova-api10:23
eglynn-office(using redis as the distributed cache IIRC)10:23
* eglynn-office wonders what happened to that effort ...10:23
shardyeglynn-office: aha, thanks, I'll see if I can track down any details10:23
eglynn-officeshardy: https://pypi.python.org/pypi/turnstile ... last updated 2013-05-01 :(10:24
shardyeglynn-office: Hmm, ok thanks, I may defer the rate-limit problem for now then..10:26
*** alexpilotti has joined #openstack-ceilometer10:26
openstackgerritAjaya Agrawal proposed a change to openstack/ceilometer-specs: Add spec for MagnetoDB metering support  https://review.openstack.org/12633510:33
cdentsigh, this identity theft saga has taken a turn for the unpleasant:10:45
cdentlast week someone at my previous address signed, in my name, to accept the delivery of a phone "I" ordered10:45
cdentnobody lives tehre10:46
*** eglynn has joined #openstack-ceilometer10:47
*** admin0 has quit IRC10:47
*** admin0 has joined #openstack-ceilometer10:50
*** safchain has quit IRC10:54
*** nosnos has quit IRC11:17
*** ildikov has quit IRC11:32
*** ildikov has joined #openstack-ceilometer11:47
*** safchain has joined #openstack-ceilometer11:47
*** ifarkas has quit IRC12:11
*** ifarkas has joined #openstack-ceilometer12:13
*** boris-42 has quit IRC12:14
*** _nadya_ has quit IRC12:18
eglynn-officeDinaBelova: FYI I've started fleshing out the influx side of our design session on "mapping gnocchi semantics onto ..."12:24
DinaBelovaeglynn-office, a-ha, cool!12:26
*** julim has joined #openstack-ceilometer12:26
eglynn-officeDinaBelova: BTW no response yet from pauldix on whether he'll be attending summit12:26
eglynn-office... which leads me to suspect he won't be12:27
DinaBelovaeglynn-office, well, it looks so..12:28
DinaBelovasad really12:28
DinaBelovahis help was quite a nice thing on the mid-cycle event12:28
*** ildikov has quit IRC12:29
eglynn-officeDinaBelova: yeap12:31
*** isviridov is now known as isviridov_away12:34
*** ajayaa has quit IRC12:34
*** janisg has quit IRC12:35
*** janisg has joined #openstack-ceilometer12:45
*** ccrouch has joined #openstack-ceilometer12:45
*** jaypipes has quit IRC12:46
*** jaypipes has joined #openstack-ceilometer12:50
*** zqfan has joined #openstack-ceilometer12:58
*** IvanBerezovskiy has quit IRC13:12
*** ddieterl_ has quit IRC13:13
*** zqfan has quit IRC13:13
*** zqfan has joined #openstack-ceilometer13:25
*** boris-42 has joined #openstack-ceilometer13:38
*** ddieterly has joined #openstack-ceilometer13:39
*** IvanBerezovskiy has joined #openstack-ceilometer13:40
*** yasu_ has quit IRC13:40
*** _nadya_ has joined #openstack-ceilometer13:40
*** _nadya_ has quit IRC13:40
*** _nadya_ has joined #openstack-ceilometer13:41
*** gordc has joined #openstack-ceilometer13:43
*** rbak has joined #openstack-ceilometer14:05
*** eglynn-office is now known as eglynn-officelun14:21
*** eglynn-officelun is now known as eglynn-off-lunch14:21
*** eglynn is now known as eglynn-lunch14:22
*** claudiub has joined #openstack-ceilometer14:38
*** edmondsw has joined #openstack-ceilometer14:51
*** boris-42 has quit IRC14:52
*** boris-42 has joined #openstack-ceilometer14:54
*** claudiub has quit IRC15:10
*** boris-42 has quit IRC15:11
*** eglynn-off-lunch is now known as eglynn-office15:12
vrovacheveglynn-office: hi, resource id on metric 'identity.authenticate.success' ( http://docs.openstack.org/developer/ceilometer/measurements.html#identity-keystone) not equal 'user ID'. It's bug or feature? ;)15:19
eglynn-officevrovachev: https://github.com/openstack/ceilometer/blob/master/ceilometer/identity/notifications.py#L57 uses the resource_info in the notification payload15:27
*** k4n0 has quit IRC15:27
eglynn-officevrovachev: ... set here by keystone https://github.com/openstack/keystone/blob/master/keystone/notifications.py#L25115:27
eglynn-officevrovachev: ... that value are you seeing?15:27
vrovacheveglynn-office: e.g. http://paste.openstack.org/show/127660/15:28
eglynn-officevrovachev: ... should we using the notification payload ['initiator']['id'] instead in that case too?15:28
eglynn-officevrovachev: openstack:8d8978d6-e71c-4016-888c-0de43f16e027 ... what's the UUID in that case?15:28
vrovacheveglynn-office: maybe, but this id not equal to any uuid for users :)15:30
vrovacheveglynn-office: user list: http://paste.openstack.org/show/127661/15:30
*** admin0 has quit IRC15:31
eglynn-officevrovachev: a tenant UUID perhaps?15:31
vrovacheveglynn-office: one moment15:31
vrovacheveglynn-office: http://paste.openstack.org/show/127662/15:32
*** admin0 has joined #openstack-ceilometer15:33
vrovacheveglynn-office: now spend sinkup on team, half an hour i will look rabbit keystone messages.15:33
*** nellysmitt has quit IRC15:34
eglynn-officevrovachev: cool15:35
*** jmatthews has joined #openstack-ceilometer15:40
vrovacheveglynn-office: Incidentally, by default Keystone notifications on devstack is turned off :) Here is my localrc: http://paste.openstack.org/show/115325/15:44
*** nellysmitt has joined #openstack-ceilometer15:44
*** nellysmitt has quit IRC15:44
*** admin0 has quit IRC15:47
*** admin0 has joined #openstack-ceilometer15:49
*** admin0 has quit IRC15:50
eglynn-officevrovachev: I suspect that a pycadf artifact15:53
eglynn-officevrovachev: i.e. the 'openstack:UUID'15:53
eglynn-officevrovachev: keystone uses pycadf to generate the events15:53
eglynn-officevrovachev: I wouldn't surprised is that mysterious UUID was generated with in pycadf15:54
eglynn-officevrovachev: e.g. here https://github.com/openstack/pycadf/blob/master/pycadf/event.py#L13515:54
vrovacheveglynn-office: heh, clearly15:55
vrovacheveglynn-office: We must make sure that keystone sends wrong messages to rabbit and create a bug. I now do it.15:57
eglynn-officevrovachev: I wonder should keystone be setting the id kwarg in the call to the pycadf eventfactory15:58
eglynn-office... https://github.com/openstack/keystone/blob/master/keystone/notifications.py#L43015:58
eglynn-officei.e. to avoid the uuid being internall generated?15:58
vrovachevhm16:00
*** _cjones_ has joined #openstack-ceilometer16:02
eglynn-officevrovachev: can you log the entire notification payload you're seeing?16:05
vrovacheveglynn-office: i have 2 environments, 1 - devstack in master(keystone notifications not works, I'm trying to understand - why), 2 mirantis openstack(with ceilo rc2 version, now I will look here)16:09
*** ifarkas has quit IRC16:10
vrovacheveglynn-office: I took it: http://paste.openstack.org/show/127678/16:21
vrovacheveglynn-office: really, id in payload not equal to any user or tenant id16:23
*** IvanBerezovskiy has left #openstack-ceilometer16:32
*** nellysmitt has joined #openstack-ceilometer16:34
*** nijaba has quit IRC16:37
*** nijaba has joined #openstack-ceilometer16:37
*** nijaba has quit IRC16:37
*** nijaba has joined #openstack-ceilometer16:37
*** edmondsw has quit IRC16:44
vrovacheveglynn-office: and i took notification to rabbit on master devstack: http://paste.openstack.org/show/127700/16:47
*** safchain has quit IRC16:56
*** rbak_ has joined #openstack-ceilometer17:00
*** rbak has quit IRC17:00
*** sbfox has joined #openstack-ceilometer17:00
*** sbfox has quit IRC17:01
*** alexpilotti has quit IRC17:04
*** zqfan has quit IRC17:06
*** harlowja_away is now known as harlowja17:08
*** bklei has joined #openstack-ceilometer17:08
*** vrovachev has quit IRC17:28
*** _nadya_ has quit IRC17:38
*** ccrouch has left #openstack-ceilometer17:41
*** cdent has quit IRC17:57
*** bklei has quit IRC18:20
*** bklei has joined #openstack-ceilometer18:21
*** sdake has joined #openstack-ceilometer18:27
*** sdake has joined #openstack-ceilometer18:27
*** eglynn-lunch has quit IRC18:34
*** eglynn-office is now known as eglynn-officeafk18:35
*** bklei has quit IRC18:45
*** 64MAAXPT5 has joined #openstack-ceilometer18:46
*** harlowja is now known as harlowja_away18:56
*** alexpilotti has joined #openstack-ceilometer19:03
*** harlowja_away is now known as harlowja19:05
*** _cjones_ has quit IRC19:11
*** alexpilotti_ has joined #openstack-ceilometer19:12
*** alexpilotti has quit IRC19:13
*** alexpilotti_ is now known as alexpilotti19:13
*** sbfox has joined #openstack-ceilometer19:19
*** lsmola_ has quit IRC19:23
*** ildikov has joined #openstack-ceilometer19:30
*** harlowja has quit IRC19:32
*** harlowja has joined #openstack-ceilometer19:35
*** eglynn-lunch has joined #openstack-ceilometer19:38
*** _nadya_ has joined #openstack-ceilometer19:42
*** sdake has quit IRC19:45
*** ildikov has quit IRC19:49
*** _cjones_ has joined #openstack-ceilometer19:56
*** srsakhamuri has joined #openstack-ceilometer19:57
*** _nadya_ has quit IRC20:01
*** lsmola has joined #openstack-ceilometer20:07
*** amalagon has joined #openstack-ceilometer20:16
*** _nadya_ has joined #openstack-ceilometer20:22
*** gordc has quit IRC20:56
*** eglynn-lunch has quit IRC21:01
*** ryanpetrello has joined #openstack-ceilometer21:09
*** 64MAAXPT5 has quit IRC21:11
*** lsmola has quit IRC21:16
*** eglynn-lunch has joined #openstack-ceilometer21:17
*** ccrouch has joined #openstack-ceilometer21:34
*** _nadya_ has quit IRC21:36
*** harlowja is now known as harlowja_away21:36
*** ryanpetrello has quit IRC21:38
*** ryanpetrello has joined #openstack-ceilometer21:42
*** sbfox has quit IRC21:48
*** harlowja_away is now known as harlowja21:51
*** harlowja is now known as harlowja_away21:51
*** harlowja_away is now known as harlowja21:51
*** boris-42 has joined #openstack-ceilometer21:55
*** nellysmitt has quit IRC22:05
*** jmatthews has quit IRC22:46
*** amalagon has quit IRC22:57
*** jmanko has quit IRC23:01
*** ddieterly has quit IRC23:03
*** ryanpetrello has quit IRC23:06
*** ruhe has left #openstack-ceilometer23:18
*** _cjones_ has quit IRC23:20
*** _cjones_ has joined #openstack-ceilometer23:21
*** _cjones_ has quit IRC23:25
*** _cjones_ has joined #openstack-ceilometer23:36
*** ccrouch has quit IRC23:36
*** ccrouch has joined #openstack-ceilometer23:41
*** rwsu has quit IRC23:45
*** ccrouch has quit IRC23:53

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