Thursday, 2015-08-06

*** r-daneel has quit IRC00:16
*** boris-42 has quit IRC00:20
*** _cjones_ has quit IRC00:35
*** thorst has joined #openstack-ceilometer00:37
*** thorst has quit IRC00:37
*** pauldix has joined #openstack-ceilometer00:54
*** pauldix has quit IRC00:59
*** pauldix has joined #openstack-ceilometer01:06
*** adurbin_ has quit IRC01:13
*** jdandrea has quit IRC01:39
*** pauldix has quit IRC01:44
*** pauldix has joined #openstack-ceilometer01:46
*** pauldix has quit IRC01:49
*** pauldix has joined #openstack-ceilometer01:50
openstackgerritRohit Jaiswal proposed openstack/ceilometer: Fix for rgw still throwing errors
*** pauldix has quit IRC01:59
openstackgerritMerged openstack/aodh: Drop downgrade field in alembic
*** logan2 has quit IRC02:08
*** logan2 has joined #openstack-ceilometer02:09
*** pauldix has joined #openstack-ceilometer02:21
*** pauldix has quit IRC03:14
*** logan2 has quit IRC03:14
*** csd has quit IRC03:27
*** boris-42 has joined #openstack-ceilometer03:27
*** csd has joined #openstack-ceilometer03:29
*** rjaiswal has quit IRC03:30
*** logan2 has joined #openstack-ceilometer03:31
*** yprokule has joined #openstack-ceilometer04:30
*** nadya has joined #openstack-ceilometer04:54
openstackgerritNadya Shakhat proposed openstack/ceilometer: Initial separating unit and functional tests
*** ankita_wagh has joined #openstack-ceilometer05:02
*** pauldix has joined #openstack-ceilometer05:14
*** pauldix has quit IRC05:19
*** jkraj has joined #openstack-ceilometer05:21
*** lsmola has joined #openstack-ceilometer05:43
*** nadya has quit IRC05:50
*** nadya has joined #openstack-ceilometer06:09
*** tasdomas has quit IRC06:17
*** boris-42 has quit IRC06:20
*** nadya has quit IRC06:32
openstackgerritOpenStack Proposal Bot proposed openstack/ceilometer: Imported Translations from Transifex
*** afazekas_ has joined #openstack-ceilometer06:37
*** belmoreira has joined #openstack-ceilometer06:52
openstackgerritMehdi Abaakouk (sileht) proposed openstack/ceilometer: integration: use the right user in gate
*** ankita_wagh has quit IRC07:04
*** ildikov has quit IRC07:07
*** nadya has joined #openstack-ceilometer07:07
*** ildikov has joined #openstack-ceilometer07:26
openstackgerritliusheng proposed openstack/ceilometer-specs: Improve Nova Instance metering
openstackgerritMerged openstack/ceilometer: Initial separating unit and functional tests
*** safchain has joined #openstack-ceilometer07:57
*** tshtilma has joined #openstack-ceilometer08:04
*** cdent has joined #openstack-ceilometer08:13
*** yassine_ has joined #openstack-ceilometer08:14
*** Ala has joined #openstack-ceilometer08:23
*** tasdomas has joined #openstack-ceilometer08:23
openstackgerritChris Dent proposed openstack/ceilometer: integration: use the right user in gate
silehtcdent, thanks :)08:32
* cdent bows08:32
openstackgerritChris Dent proposed openstack/gnocchi: Add to influx command's $PATH
*** tshtilma has quit IRC08:54
*** jkraj has quit IRC09:04
*** bdossant has joined #openstack-ceilometer09:25
openstackgerritChris Dent proposed openstack/ceilometer: Sync devstack plugin with devstack:lib/ceilometer
*** jkraj has joined #openstack-ceilometer09:40
openstackgerritMehdi Abaakouk (sileht) proposed openstack/ceilometer: Fixes the kafka publisher
openstackgerritMehdi Abaakouk (sileht) proposed openstack/ceilometer: api: fix alarm deletion and update
*** yassine_ has quit IRC10:03
*** nadya has quit IRC10:06
cdentsileht: a new gabbi just released with your changes and also the query_parameters thing10:14
silehtcdent, cool thanks10:14
openstackgerritMehdi Abaakouk (sileht) proposed openstack/ceilometer: Enable entry points for new declarative meters
openstackgerritMehdi Abaakouk (sileht) proposed openstack/ceilometer: Convert identity, sahara and volume to meters yaml
openstackgerritMerged openstack/ceilometer: integration: use the right user in gate
openstackgerritChris Dent proposed openstack/aodh: Correct database functional tests
*** jkraj has quit IRC11:01
cdentjd__: we wanna call the influxdb + metricd poor interaction a bug for later?11:05
jd__cdent: nah I'm gonna fix it in the patch I think it's important11:06
cdentah okay, roger11:06
jd__who's that Roger?11:07
* jd__ hides11:07
*** liusheng has quit IRC11:08
openstackgerritIgor Degtiarov proposed openstack/ceilometer: Cast Int64 values to float
*** liusheng_ has joined #openstack-ceilometer11:10
openstackgerritIgor Degtiarov proposed openstack/ceilometer: Cast Int64 values to float
openstackgerritMerged openstack/aodh: gnocchi: percent_of_overlap=0 for agg. alarms
*** pauldix has joined #openstack-ceilometer11:17
*** jkraj has joined #openstack-ceilometer11:19
*** pauldix has quit IRC11:22
openstackgerritMerged openstack/ceilometer: Drop out-of-time-sequence rate of change samples
*** thorst has joined #openstack-ceilometer11:24
openstackgerritMerged openstack/aodh: Delete its corresponding history data when deleting an alarm
*** gordc has joined #openstack-ceilometer11:28
openstackgerritMerged openstack/ceilometer: Sync devstack plugin with devstack:lib/ceilometer
*** nadya has joined #openstack-ceilometer11:35
openstackgerritMerged openstack/gnocchi: rest: directly pass metric to search_value()
openstackgerritMerged openstack/gnocchi: indexer: always eagerly load archive_policy
cdentwe move a lot of code lately11:38
cdentI hope it all doesn't blow up in our face11:39
gordccdent: which code?11:39
cdentgordc: in general across all three of ceilo, gnocchi, aodh: lots of code being written and reviewed and merged11:40
cdentand that's a) awesome, go us b) a bit scary because there are holes in our testing11:40
gordccdent: well that's why we kept alarm code in ceilometer. in case "all goes to sh*t"-scenario11:41
gordci'm just thankful we didn't do the poll/notification agent split11:41
*** yassine_ has joined #openstack-ceilometer11:42
openstackgerritgordon chung proposed openstack/ceilometer: Revert "Revert "remove instance:<flavor> meter""
cdentgordc: clearly we have different approaches11:45
*** pauldix has joined #openstack-ceilometer11:45
cdentI think we want stuff to go to shit because it is already shit and by making it visible we are more likely to fix it11:45
cdentthe crux is making sure we actually marshal ourselve appropriately11:46
gordci'm not sure you're necessarily making existing sh*t visible.11:46
openstackgerritgordon chung proposed openstack/ceilometer: add flexible grouping key
cdentmaybe i have big eyes11:48
openstackgerritgordon chung proposed openstack/ceilometer: group pollsters by interval
openstackgerritgordon chung proposed openstack/ceilometer: add poll history to avoid duplicate samples
openstackgerritgordon chung proposed openstack/ceilometer: disable non-metric meter definitions
openstackgerritChris Dent proposed openstack/aodh: Correct database functional tests
*** jdandrea has joined #openstack-ceilometer12:30
*** llu has quit IRC12:32
*** edmondsw has joined #openstack-ceilometer12:36
*** jasonamyers has joined #openstack-ceilometer12:41
*** afazekas_ has quit IRC12:50
jd__cdent: AFAICS metricd with InfluxDB will just do nothing and wait :13:03
cdentand not spew errors. I guess that abstract base class magic is more magic than I knew. It still seems wrong that it would even run13:04
*** safchain has quit IRC13:04
cdenti'd prefer to see it quit with a message13:05
gordcprad: i might have a solution for multi-meter13:14
gordcwe can use jsonpath. if the user defines path rather than static string we will find it and built a meter for each match.13:15
*** yprokule has quit IRC13:16
*** yprokule has joined #openstack-ceilometer13:18
*** thumpba has joined #openstack-ceilometer13:23
openstackgerritJulien Danjou proposed openstack/gnocchi: Add to influx command's $PATH
openstackgerritJulien Danjou proposed openstack/gnocchi: InfluxDB storage driver
jd__cdent: hmmm the thing is that I'm afraid that if it just exists, people will tend to think it's not working because the ref doc will say "it has to run"13:25
jd__cdent: s/exists/exits/13:25
cdentexit with a message or something is what I meant13:26
*** thumpba_ has joined #openstack-ceilometer13:26
cdentIt's bad mojo to run 8 processes of a long running daemon that is in a tight loop but doing nothing13:26
gordccan't you add note it's only needed for canonical driver?13:26
cdentgordc: even then if you accidently start it up it shouldn't stay up if it is not doing anything13:27
gordccdent: can you tell it's doing nothing?13:28
gordctell if it's*13:29
cdentwhy does that matter? or: What are you actually asking13:29
*** thumpba has quit IRC13:30
gordcwell you can start a collector in ceilometer but if you're publishing to another queue, it'll do nothing.13:30
*** thumpba_ has quit IRC13:30
cdentthat's pretty weak sauce too, just because sets a bad precendent doesn't mean gnocchi should follow it?13:30
gordcoh i'm just asking how tightly coupled they are... i'm not criticising your answer13:31
gordccollector and notification are meant to be independant... there could be a perfectly valid reason one exists but not the other.13:32
gordcor both exist but unrelated13:32
*** safchain has joined #openstack-ceilometer13:32
jd__cdent: "exit with a message" this is so 199913:33
jd__no devops deploying over Puppet or anything like that is going to read that :)13:34
jd__it's just going to think "service crassshheeed"13:34
jd__but yeah having 8 process doing nothing isn't really friendly13:34
cdentjd__: the devop should never deploy it in the first place13:34
* jd__ high fives cdent 13:34
cdentI'm just talking about the person who runs it without knowing what they are doing13:34
cdent(which is a lot of people)13:34
*** thumpba has joined #openstack-ceilometer13:36
jd__I wonder if we need to do something async with continuous query?13:37
* cdent is too ignorant about influxdb to comment13:39
*** Ala has quit IRC13:39
*** edmondsw has quit IRC13:46
*** Ala has joined #openstack-ceilometer13:47
gordcnadya: i think i asked this earlier, but are you (or someone) planning to work on adding postgres tests to gate.13:47
nadyagordc: yep, I think we may start working on this13:52
gordcnadya: cool cool. i'll leave it to you folks then. :)13:53
nadyagordc: ok  :)13:53
gordcthanks for working on the functional tests. nice to have the split in.13:54
gordcsileht: "add option to disable metering apis when gnocchi is enabled or create wsgi pipeline.13:55
gordcthat should be done right?13:55
gordcit's from
silehtgordc, yes this is done13:56
gordcawesome :)13:56
nadyagordc: it was for too long tbh13:56
gordcnadya: yeah. i think it's something that needs to be ready after RC so it can be merged quickly/easily.13:57
gordcbut glad it's in now regardless.13:57
openstackgerritJulien Danjou proposed openstack/gnocchi: Add to influx command's $PATH
openstackgerritJulien Danjou proposed openstack/gnocchi: InfluxDB storage driver
nadyagordc: I hope we will speed up now :)13:59
gordcnadya: that's not opensource...14:00
* gordc dodges bricks14:00
*** r-daneel has joined #openstack-ceilometer14:01
*** rbak has joined #openstack-ceilometer14:02
cdentgordc I'm going to have to run against you for PTL. Not because I want it but because you are a heretic and need to be deposed.14:05
gordclol fair enough14:13
*** afazekas_ has joined #openstack-ceilometer14:27
*** edmondsw has joined #openstack-ceilometer14:28
openstackgerritPradeep Kilambi proposed openstack/ceilometer: Enable entry points for new declarative meters
openstackgerritPradeep Kilambi proposed openstack/ceilometer: Convert identity, sahara and volume to meters yaml
*** pradk has joined #openstack-ceilometer14:34
pradksileht, the tests were failing on the dependent patches .. so rebased again, hope this time it goes through.. if you can fast +A again please14:37
gordcpradk: we need eyes on the devstack patch right?14:38
pradkyea we have a few +1's, need +2's14:38
silehtpradk, I have tried to save people times this morning, because all already approved patched was failed due to the unit/functional tests split merge14:39
pradkunderstood, thx sileht14:39
pradksileht, you have devstack powers?14:39
gordci put a call out on infra14:39
gordci should mention that i have bad cred over there.lol14:40
gordcdamn. got to repost on qa... i'm out. it'll probably be bad if i post there.lol14:42
*** kfarhane_ has joined #openstack-ceilometer14:42
kfarhane_please can someone explain to me, why couldn't see cpu_util metric for resources in compute node?14:43
pradkhehe gordc14:43
*** afazekas_ has quit IRC14:43
gordckfarhane_: which hypervisor?14:45
*** MVenesio has joined #openstack-ceilometer14:45
kfarhane_gordc kvm hypervisor in a compute node14:47
gordckfarhane_: do you have cpu meters?14:48
kfarhane_gordc: neither14:49
gordckfarhane_: ok. because the cpu_util meter is derived from cpu meter (depending on libvirt version)14:49
*** broksos has joined #openstack-ceilometer14:49
gordckfarhane_: do you have any errors in compute agent log?14:50
kfarhane_gordc: i explain more, for instance in controller node, i have theses meters but no for instance in compute nodes14:50
gordcyou have nova on controller node?14:51
kfarhane_gordc: compute agent is running whitout errors, i can see : "Polling pollster cpu_util in the context of meter_source" in the screen of ceilometer compute agent,14:52
broksosIm a bit stumped on an event query via the v2 api.  I am collecting events, can see them with event-list, all is fine.14:52
broksosI can query them by event_type without issue14:52
broksosHowever, I need to query them by date ie; all events since datetime14:52
kfarhane_in my compute node, i activate only ceilometer-acompute, and other service of ceilomter are activated in controller14:53
broksosthis does not return any results..  I've looked through the code and the database... the model returns this field as 'generated' which is fine14:53
broksosbut searching on it doesn't seem to work... when I look through the code it seems to want to treat all but these 4 things as traits:14:54
gordckfarhane_: that makes sense... if you don't have cmpute agent on controller node, i don't see how you'd get cpu_util meters from there.14:55
broksos    evt_model_filter = {         'event_type': None,         'message_id': None,         'start_timestamp': None,         'end_timestamp': None     }14:55
broksosIf I try and search by start_timestamp, end_timestamp or generated all return empty14:55
broksosany ideas on what I am missing?14:55
kfarhane_gordc: there is a compute agent in controller node, and in compute node too,14:56
kfarhane_i think compute agent will retrieve data from pollster and send them to the collector which is in the controller14:57
gordckfarhane_: which version are you?14:57
gordckfarhane_: broksos: we're about to have ceilometer meeting in 2 minutes. will get back to you afterwards.14:58
gordcfeel free to open bug or ask on mailing list if you need to drop.14:58
*** r-daneel has quit IRC14:58
*** jkraj has quit IRC14:58
broksosgordc: thank you14:58
silehtcdent, I have talked with jamielennox about my keystonemiddleware patch and it's not fan of my change14:59
cdentis there an alternative?14:59
cdentso I went ahead and changed the devstack plugin to use the paste file14:59
silehtjust put the options in paste-deploy config and loose all oslo.config features15:00
kfarhane_do you mean version of devstack? i'm in version stable/juno15:00
cdent(in that patchset I've been working on for what seems like forever now, but _may_ be finally there)15:00
silehtcdent, I will start a thread on the ML15:00
cdentseems like a good idea15:00
cdentyou've seen the other bug too, yeah?15:00
silehtcdent, many other middlewares have the same issues15:00
silehtcors, ssl, sizelimit.15:00
*** eglynn has joined #openstack-ceilometer15:01
*** r-daneel has joined #openstack-ceilometer15:01
silehtcdent, I have learn that zaqar do the monkey patch of auth_token.CONF like you do15:01
* sileht cries15:01
cdentthat seems _really_ dirty15:01
cdentI've abandoned that15:01
silehtI have started to write the mail, so wait and see15:01
* cdent needs a new clock15:04
openstackgerritMerged openstack/ceilometer: Imported Translations from Transifex
openstackgerritMerged openstack/ceilometer: Fixes the kafka publisher
*** Ala has quit IRC15:14
*** vishwanathj has joined #openstack-ceilometer15:19
*** llu-laptop has joined #openstack-ceilometer15:22
*** llu-laptop has left #openstack-ceilometer15:22
*** pauldix has quit IRC15:27
*** pauldix has joined #openstack-ceilometer15:28
gordcdhellmann: just a headsup, we'll need a release for ceilometerclient next week (seems too late in week)15:30
*** liusheng_ has quit IRC15:35
*** tongli has joined #openstack-ceilometer15:44
*** _cjones_ has joined #openstack-ceilometer15:55
cdentjason was hoping we'd address or push forward the question of the mid-cycle16:00
silehtcdent, I have sent the mail to ML16:00
*** shang has joined #openstack-ceilometer16:01
silehtit's a bit long, I hope it's understandable16:01
gordceglynn: if we keep it for liberty. the packagers won't be using it (assuming aodh is good).16:01
gordceglynn: so i don't think it even needs to be there? do ppl use ceilometer-code not through packages?16:02
gordceglynn: the idea of keep it till M* was as a fallback... apparently that fallback isn't actually valid.16:02
pradki think the fallback was to give packagers more time16:03
gordccdent: yeah i was going to bring it up at open discussion... but more important matters.16:03
pradkaodh doesnt have a release yet16:03
*** belmoreira has quit IRC16:03
pradkso we cant officially get it into repos unless its on pypi or somewhere official16:04
pradkassuming that happens we should be good16:04
*** shang has left #openstack-ceilometer16:04
ildikovI'm a bit uncomfortable to drop it a bit out of the blue, but considering that we supposed to have new features added to Aodh only we need it in a package soon anyway I guess we will really focus on getting it released in a reasonable timeframe16:04
ildikovdo we know if there is anything blocking that to happen?16:05
gordci haven't seen any blockers... we just need to gate on it to verify.16:05
ildikovgordc: cool, that sounds good so far16:06
gordcall the current patches seems to be new features16:06
cdentdo we have confidence that our testing is actually good enough to verify aodh is "healthy". If not, do we have alternate means of achieving the  same thing?16:06
gordcthe existing integration tests continue to work against aodh? ie heat won't break.16:08
gordceglynn: ^ that's the original note we wrote right after midcycle16:08
ildikovgordc: we have a "fix" in the client to keep Heat happy, right?16:09
gordcildikov: there's redirects in both ceilometer-api and client that will point to aodh if configured.16:10
gordcnot tested in gate16:10
ildikovgordc: yeap, this is what I meant16:10
*** yassine_ has quit IRC16:12
gordccdent: i'm not familiar with all the methods ppl have to use ceilometer. it seems from package pov, they both can't coexist.16:12
ildikovgordc: what is missing for testing it on the gate?16:12
gordcwhat other ways are there?16:12
gordcildikov: sileht  is working on that currently.16:12
ildikovgordc: ok, cool16:13
cdentthere's a weird bug that is making both ceilometer and aodh tests rather unreliable at the moment:
openstackLaunchpad bug 1481244 in Ceilometer "TranslationHook middleware can lose track of its thread local storage" [Undecided,New]16:13
gordcseems he has a workaround to blocker but if we can gate and ensure heat ain't broke i think we should be ok.16:13
cdentthat's causing this to never quite work as well on the experimental jobs as it should16:14
cdentthat patch allows the aodh tests to actually run instead of alway skipping16:14
cdentthere were a _lot_ of problems16:14
cdentI don't know how to address the 1481244 bug, anyone have ideas?16:15
cdent(and unfortunately it is just masking some other bug)16:15
gordcwhere's the bug showing up in your patch?16:15
gordccdent: it happens in both aodh and ceilometer but not when you use eventlet api?16:18
cdentI'm not certain, generally I run mod wsgi where possible, aodh will follow whatever ENABLE_HTTPD_MOD_WSGI_SERVICES16:20
* gordc never runs mod wsgi.16:21
cdentyou are totally a grape in the path of the steamroller or progress16:21
cdentwe should be running nginx+uwsgi by now...16:21
openstackgerritRohit Jaiswal proposed openstack/aodh: Fixes querying alarm history with severity field
gordctoo damn hard to debug multiple threads.16:22
gordcyou use your fancy stuff for production.16:22
*** idegtiarov has quit IRC16:22
*** rjaiswal has joined #openstack-ceilometer16:22
*** exploreshaifali has joined #openstack-ceilometer16:24
*** safchain has quit IRC16:26
gordccdent: yeah, i don't see anything different between the two16:27
*** kbyrne has quit IRC16:29
cdentgordc: i'm used up for the evening so can't look any more today. It's not blocking that change but it is blocking the usefulness of the functional jobs for aodh16:30
cdent(still experimental)16:30
gordccdent: kk. will look and review your grenade stuff for now16:30
gordcbroksos: start_timestamp and end_timestamp will query against generated attr16:32
* cdent heads16:33
*** cdent has quit IRC16:33
*** kbyrne has joined #openstack-ceilometer16:33
*** nadya has quit IRC16:37
*** ildikov has quit IRC16:38
*** jdandrea has quit IRC16:38
*** jdandrea has joined #openstack-ceilometer16:39
openstackgerritRohit Jaiswal proposed openstack/ceilometer: Fix for rgw still throwing errors
broksosgordc: ok - I must be doing it wrong :) - Let me poke at it a bit more, thanks!16:40
*** jmank has quit IRC16:41
*** bdossant has quit IRC16:41
gordcbroksos: you'll probably want something in format like: -q 'start_timestamp=2013-10-01T09:00:00'16:42
broksosgordc: this returns empty:16:50
broksosceilometer event-list -q 'start_timestamp=>2013-10-01T09:00:00'16:50
broksosI've also been using the client from python with this sort of thing:16:51
broksosmyq=[{"field": "start_timestamp",      "op": "ge",      "type": "datetime",      "value": "2015-08-04T15:08:15.737000"},     {"field": "end_timestamp",      "op": "le",      "type": "datetime",      "value": "2015-08-07T15:08:15.737000"}]16:51
gordcbroksos: it only does equivalence.16:51
gordcstart_timestamp=2013-10-01T09:00:00 will give you everything after that time16:51
broksossorry, started with that, it's just as empty16:52
gordcbroksos: which backend? and build?16:52
gordci'll give it a try.16:53
gordcalso, you need to be admin role to view events.16:53
*** ankita_wagh has joined #openstack-ceilometer16:55
broksosmongo, v2 api, openstack-ceilometer-api-2014.2.3-5.el7ost.noarch, python-ceilometerclient-1.0.12-1.el7ost.noarch16:55
broksosgordc: thanks - it returns events fine without a filter16:55
broksosor I can set the filter to event_type=  and use one of those and it's fine16:56
broksosit's just dates that are not right for me16:56
gordcdarn. my live setup is sql... i'll need to setup mongo.16:57
*** yprokule has quit IRC16:58
gordccan you try start_time instead of start_timestamp?16:59
gordcbroksos: ^16:59
broksosgordc: ack - checking17:00
*** htruta_ has quit IRC17:00
*** htruta has joined #openstack-ceilometer17:01
broksosgordc: winning!  Thanks for having a look17:01
broksosmaybe I'll file/fix a doc bug :)17:01
gordcbroksos: cool cool. yeah i think we sanitised the names to start_timestamp in kilo becuase they were different from sample query param17:02
gordcbroksos: sounds good.17:02
gordcthanks for filing bug17:02
eglynngordc: sorry, I had to drop off the aodh discussion earlier as I had another call17:10
eglynngordc: ... did it converge on a conclusion?17:10
*** ildikov has joined #openstack-ceilometer17:15
*** rbak has quit IRC17:15
gordceglynn: i don't believe we found a use case that didn't involve going through packages... if packages is the only point of consumption then the mailing list note remains true. "ceilo-alarm alarm will be deleted when aodh is working" working="gated, released, packaged"17:16
gordcif that isn't the case, only bug fixes will be backported to remaining ceilo-alarm codebase.17:17
eglynngordc: sounds reasonable to me ... and presumably the aodh release cadence is decoupled from liberty, so aodh could conceivably be released prior to the "integrated" liberty release is tagged?17:19
eglynn*being tagged17:19
gordceglynn. yes. that's what i'm hoping. haven't experience how easy it is for libraries on independant release cycles.. but that is the plan.17:23
gordcif not, we may need to settle for ghost code in ceilometer repo17:23
*** rbak has joined #openstack-ceilometer17:27
eglynngordc: cool17:28
eglynngordc: I'll stick a note on the o.vo review17:30
gordceglynn: kk.17:30
gordclet me know if you spot any issues in plan... i think we're all pretty new to this split business.17:31
*** srsakhamuri has joined #openstack-ceilometer17:38
eglynngordc: I've updated with a summary of the conclusion ... please feel free to correct me if I went off-base on that17:39
*** alexpilotti has joined #openstack-ceilometer17:42
*** rbak has quit IRC17:42
*** exploreshaifali has quit IRC17:47
*** exploreshaifali has joined #openstack-ceilometer17:52
openstackgerritJulien Danjou proposed openstack/aodh: devstack: use $API_WORKERS to set the number of WSGI workers in Apache
openstackgerritJulien Danjou proposed openstack/aodh: devstack: only use 1 thread
*** exploreshaifali has quit IRC17:54
*** thumpba has quit IRC17:54
*** thumpba has joined #openstack-ceilometer17:57
openstackgerritJulien Danjou proposed openstack/gnocchi: devstack: use $API_WORKERS to set the number of Apache WSGI workers
pradkgordc, so our discussion from yesterday on swiftmiddleware meters - the issue is the metric list18:00
pradkgordc, i could do something like but 0 or 1 as the index is a hit or a miss18:01
gordcpradk: i'm thinking we can use jsonpath to actually query for it.18:03
gordcthe idea was maybe we allow name to be a path and not just a static string18:04
pradkyea the jsonpath_rw does go by path which is what we do today.. but it cant handle lists in the path18:05
pradkit expects a nested dict to query18:05
gordcso for swiftmiddleware, we can query for payload[measurements][*][metric]18:05
gordcyou can query for array
pradkhmm lemme try i doubt it18:06
gordci think it'd be payload[measurements][*][metric][name] to be exact18:07
gordcit would return a list18:08
gordcand based on that, you can decide how many samples to build18:08
gordci had it partially working.. only did the metric name part.18:08
pradkarent those meter names standard? or can they change?18:09
gordclet me know if you want to split tasks.18:09
pradki remember you mentioning its always 218:09
pradkso why not just do what i did in my gist above and try the match you suggested just for volume18:10
gordcyou mean are they always present?18:10
gordcno. you may get two, one, or zero18:10
pradkah ok18:10
gordcmaybe not two... definitely one or zero18:11
gordcjd__: is that a fix or a workaround?18:13
*** thumpba_ has joined #openstack-ceilometer18:14
*** eglynn has quit IRC18:15
*** thumpba has quit IRC18:17
*** gordc_internal has joined #openstack-ceilometer18:29
openstackgerritRohit Jaiswal proposed openstack/ceilometer: Adds support for dynamic event pipeline
pradkgordc, i'm currently looking into cpu math meters, i'm happy to split a few if you want to handle the bulk meters18:52
pradkgordc, i'll get to the bulk ones next so if you dont get to it by then i'll pick it up18:53
pradkfor now i want these cpu math ones done18:53
*** harlowja has quit IRC18:56
*** nadya has joined #openstack-ceilometer18:56
gordc_internalpradk: that sounds fine. i may have some time later today/tomorrow.18:57
gordc_internaljust ask before you start working on it in case i do manage to start.18:57
pradkcool cool18:57
*** rbak has joined #openstack-ceilometer19:02
*** nadya has quit IRC19:08
*** ankita_w_ has joined #openstack-ceilometer19:17
*** ankita_w_ has quit IRC19:20
*** ankita_wagh has quit IRC19:20
*** ankita_wagh has joined #openstack-ceilometer19:20
*** thumpba_ has quit IRC19:34
*** thumpba has joined #openstack-ceilometer19:35
*** boris-42 has joined #openstack-ceilometer19:39
*** broksos has quit IRC19:39
*** jruano has joined #openstack-ceilometer19:41
*** hhuang has quit IRC19:42
*** thumpba has quit IRC19:48
*** thumpba has joined #openstack-ceilometer19:48
*** ankita_wagh has quit IRC19:51
*** thumpba has quit IRC19:53
*** thumpba has joined #openstack-ceilometer19:55
*** hhuang has joined #openstack-ceilometer19:55
*** thumpba has quit IRC19:56
*** thumpba has joined #openstack-ceilometer19:57
*** ankita_wagh has joined #openstack-ceilometer19:58
*** jruano has quit IRC20:01
*** gordc_internal has quit IRC20:10
*** wayneeseguin has quit IRC20:11
*** pradk has quit IRC20:11
*** logan2 has quit IRC20:12
*** logan2 has joined #openstack-ceilometer20:12
*** thumpba_ has joined #openstack-ceilometer20:12
*** pradk has joined #openstack-ceilometer20:12
*** wayneeseguin has joined #openstack-ceilometer20:13
*** srsakhamuri has quit IRC20:14
*** srsakhamuri has joined #openstack-ceilometer20:14
*** thumpba has quit IRC20:16
*** jasonamyers has quit IRC20:18
*** DCWilliams_VA has joined #openstack-ceilometer20:21
*** DCWilliams_VA has quit IRC20:26
*** tongli has quit IRC20:34
*** slogan621 has joined #openstack-ceilometer20:44
slogan621How do I register an agent that reponds to notifications on the OSLO bus and republishes in a way compatible with ceilometer collector? I understand the code that implements event_types and process_notifications and publishes samples,20:48
*** logan2 has quit IRC20:48
slogan621what I don't understand if that is just some service I write and run myself, or if it is something that I register with ceilometer in some way and it will load ad call me20:48
gordcslogan621: one way is to copy this: and modify it to your own notificatoin... then add it to setup.cfg undert ceilometer.notification20:51
gordcslogan621: or you can wait for which will do the same... but you just need to edit a yaml file20:51
gordcassuming this is an exchange20:52
slogan621gordc: hi! BTW I am now successfully publishing to OSLO bus, thanks for your pointers to code20:52
gordcexisting exchange.*20:52
gordcslogan621: awesome! did the middleware stuff help?20:52
gordci'll assume yes and take credit :)20:53
slogan621a bit, I think it ended up being alook at a lot of things and then coming up with something that work. I had to work through various things like oslo_config which was new to me, but essential in order to get things like rabbitmq passwords20:53
gordcah cool cool20:54
slogan621once I get my company to consent I need to blog my expperiences20:54
slogan621so much is assumed in the current documentation that you need to know, but ot called out20:55
slogan621er, not called out20:55
slogan621so, now at the stage of sniffing these publications and generating samples20:56
slogan621and republishing20:56
gordcfair enough. i think a lot of us are just so use to everything.20:56
slogan621I'm a newbie to ceilometer, much more smart on neutron and network virtualization, which I have been involved with for a couple of years20:57
gordcyou adding more neutron meters?20:58
slogan621more hardware meters (I work on network switching hardware) - stuff that complements what neutron is publishing about tenant networks already20:59
slogan621sort of researchy at this point20:59
gordci see. i'll assume you're on a separate exchange seeing as you aren't reusing neutron notification code.21:01
*** ankita_w_ has joined #openstack-ceilometer21:01
slogan621my own exchange and queue, definitely21:01
slogan621generally speaking I think the goal is that something might call ceilometer API to get my data, and get neutron data, and put the two together in a way that helps someone better understand how the reports about the virtua network are affected by the state of the physical21:03
gordcseems reasonable.21:03
*** ankita_wagh has quit IRC21:04
slogan621I'm making a bet that ceilometer is the place to put this kinda of data, that it will be the underpinings of a metering as a service component of openstack that has yet to appear21:05
*** thumpba_ has quit IRC21:08
gordccool cool. yeah. ceilometer is the telemetry component of openstack... so basically it captures state/details of openstack resouces.21:08
*** rbak has quit IRC21:08
slogan621gordc: good, looks like I'm doing the right things - publishing my data raw to OSLO, doing a notification approach versus polling, using my own exchange, and chosing to use ceilometer as the way to do all this21:10
gordc:) if you have any questions feel free to ask here... or on mailing list (most of the dev team is in euro/asia region)21:11
slogan621this review you pointed me at, will the approach used by objectstore now be deprecated once it is merged, or will the two approaches co-exist?21:11
gordci don't think it'll go away, since the general framework is the same (just the processing part is different).21:12
slogan621ah, ok21:12
slogan621well, time to look at the objectstore code - thanks again!21:13
gordcnew meters will be done the new way as it doesn't require a code change but if you code it using the code example i gave you, it (probably) won't break21:13
gordcslogan621: kk. good luck21:13
slogan621oh man that looks simple21:15
*** ankita_wagh has joined #openstack-ceilometer21:16
* gordc hopes so21:16
slogan621I had seen the code in objectstore, and it maps to what I read in the docs (implement a process_notificaton function and a get_targets, but I completely missed on the setup.cfg step.21:18
slogan621something that fixes this on contributing/plugins.html would probably unstick many others21:19
*** ankita_w_ has quit IRC21:20
gordcslogan621: makes sense. want to open a bug at and list exactly what you would want to see?21:20
gordcor you can patch it yourself :)
slogan621let me see if I can do that without getting fired :-) my company has not yet allowed me to contribute (working on it).21:21
gordcslogan621: lol21:21
gordcyeah you probably need to sign CLA and everything... bug might be easier.21:21
gordcthat or create a pseudo name21:22
slogan621yeah, that might be the right way.21:22
slogan621ok, time to code this up - I'll let you know how it goes - you have been essentially my lifeline up to this point, thanks so very much21:24
*** MVenesio has quit IRC21:24
gordcslogan621: np dude.21:24
*** ankita_wagh has quit IRC21:28
*** ankita_w_ has joined #openstack-ceilometer21:28
*** ankita_w_ has quit IRC21:32
*** ankita_wagh has joined #openstack-ceilometer21:32
*** harlowja has joined #openstack-ceilometer21:34
*** belmoreira has joined #openstack-ceilometer21:35
*** harlowja has quit IRC21:39
*** harlowja has joined #openstack-ceilometer21:39
*** thorst has quit IRC21:55
*** pradk has quit IRC22:00
*** rjaiswal has quit IRC22:00
*** gordc has quit IRC22:08
*** slogan621 has quit IRC22:09
*** ankita_wagh has quit IRC22:11
*** ankita_wagh has joined #openstack-ceilometer22:11
*** edmondsw has quit IRC22:20
*** jfluhmann has joined #openstack-ceilometer22:36
*** ankita_wagh has quit IRC22:46
*** alexpilotti has quit IRC22:52
*** Ephur has quit IRC22:57
*** alexpilotti has joined #openstack-ceilometer22:58
*** logan2 has joined #openstack-ceilometer23:06
*** ankita_wagh has joined #openstack-ceilometer23:14
*** r-daneel has quit IRC23:17
*** rjaiswal has joined #openstack-ceilometer23:18
*** drjones has joined #openstack-ceilometer23:24
*** _cjones_ has quit IRC23:25
*** drjones has quit IRC23:28
*** pauldix has quit IRC23:34
*** jfluhmann has quit IRC23:35
*** alexpilotti has quit IRC23:58

Generated by 2.14.0 by Marius Gedminas - find it at!