Thursday, 2014-09-04

*** r-daneel has quit IRC00:03
*** r-daneel has joined #openstack-ceilometer00:04
*** nikhil__1 has joined #openstack-ceilometer00:11
*** therve has joined #openstack-ceilometer00:11
*** ekarlso- has joined #openstack-ceilometer00:11
*** ifarkas_ has joined #openstack-ceilometer00:12
*** stenchlarge has quit IRC00:12
*** ekarlso has quit IRC00:12
*** therve` has quit IRC00:12
*** nikhil___ has quit IRC00:12
*** lynxbat has quit IRC00:12
*** dhellmann_ has quit IRC00:12
*** xmltok_ has quit IRC00:12
*** ifarkas has quit IRC00:12
*** lynxbat has joined #openstack-ceilometer00:12
*** dhellmann has joined #openstack-ceilometer00:12
*** claudiub has quit IRC00:13
*** xmltok has joined #openstack-ceilometer00:13
*** packet has quit IRC00:18
openstackgerritgordon chung proposed a change to openstack/python-ceilometerclient: test
openstackgerritA change was merged to openstack/ceilometer: XenAPI support: Update measurements documentation
*** lynxbat has quit IRC00:28
*** promulo_ has joined #openstack-ceilometer00:40
*** r-daneel has quit IRC00:43
*** promulo_ has quit IRC00:54
*** atan8 has joined #openstack-ceilometer00:56
*** drjones has quit IRC01:06
*** gzhai has joined #openstack-ceilometer01:06
*** _cjones_ has joined #openstack-ceilometer01:06
*** _cjones_ has quit IRC01:11
*** atan8 has quit IRC01:14
*** liusheng has joined #openstack-ceilometer01:21
*** nosnos has joined #openstack-ceilometer01:54
*** theanalyst has quit IRC02:13
*** theanalyst has joined #openstack-ceilometer02:15
*** adriant has joined #openstack-ceilometer02:22
*** lynxbat has joined #openstack-ceilometer02:25
*** lynxbat has quit IRC02:33
*** dougwig has quit IRC02:36
*** dougwig_ has joined #openstack-ceilometer02:39
*** sbfox has quit IRC02:52
*** itisit has joined #openstack-ceilometer02:58
*** tongli has quit IRC03:05
*** ajc_ has joined #openstack-ceilometer03:14
*** shardy_z has quit IRC03:40
*** _cjones_ has joined #openstack-ceilometer03:46
*** _nadya_ has joined #openstack-ceilometer04:12
*** _nadya_ has quit IRC04:17
*** _nadya_ has joined #openstack-ceilometer04:47
*** itisit has quit IRC04:50
*** _cjones_ has quit IRC05:05
*** _cjones_ has joined #openstack-ceilometer05:05
*** _nadya_ has quit IRC05:10
*** _nadya_ has joined #openstack-ceilometer05:11
*** _nadya_ has quit IRC05:17
openstackgerritgordon chung proposed a change to openstack/ceilometer: normalise resource data
*** nsaje-office is now known as nsaje05:41
*** Longgeek has joined #openstack-ceilometer05:44
*** _nadya_ has joined #openstack-ceilometer05:48
*** adriant has quit IRC05:50
*** harlowja is now known as harlowja_away05:50
*** _cjones_ has quit IRC06:00
*** _cjones_ has joined #openstack-ceilometer06:01
*** eglynn_ has joined #openstack-ceilometer06:02
*** _cjones_ has quit IRC06:05
openstackgerritOpenStack Proposal Bot proposed a change to openstack/ceilometer: Imported Translations from Transifex
*** Longgeek has quit IRC06:12
*** Longgeek has joined #openstack-ceilometer06:18
*** Longgeek_ has joined #openstack-ceilometer06:21
*** Longgeek_ has quit IRC06:22
*** Longgeek_ has joined #openstack-ceilometer06:23
*** Longgeek has quit IRC06:24
*** Longgeek has joined #openstack-ceilometer06:24
*** Longgeek_ has quit IRC06:26
*** samta has joined #openstack-ceilometer06:27
*** _nadya_ has quit IRC06:57
*** adam_g has quit IRC07:04
*** adam_g has joined #openstack-ceilometer07:06
*** adam_g has joined #openstack-ceilometer07:06
*** _nadya_ has joined #openstack-ceilometer07:16
*** Longgeek has quit IRC07:20
*** Longgeek has joined #openstack-ceilometer07:21
*** Longgeek_ has joined #openstack-ceilometer07:26
*** Longgeek_ has quit IRC07:27
*** Longgeek_ has joined #openstack-ceilometer07:27
*** Longgeek has quit IRC07:29
*** shardy has joined #openstack-ceilometer07:30
openstackgerritliusheng proposed a change to openstack/ceilometer: Check unsupported query filters of listing events
*** ildikov has joined #openstack-ceilometer07:31
*** eglynn_ has quit IRC07:31
*** Longgeek has joined #openstack-ceilometer07:33
*** Longgeek_ has quit IRC07:36
*** Longgeek has quit IRC07:37
*** Longgeek has joined #openstack-ceilometer07:38
*** mspreitz has quit IRC07:40
nsajeeglynn__: ouch, I just realized where Ironic drew inspiration for its hash ring impl:
*** gzhai has quit IRC07:48
*** arezmerita has quit IRC07:56
*** safchain has joined #openstack-ceilometer08:13
*** jbuckley has joined #openstack-ceilometer08:17
*** eglynn_ has joined #openstack-ceilometer08:31
nsajeeglynn_: ouch, I just realized where Ironic drew inspiration for its hash ring impl:
nsajeeglynn_: I tested the code from the blogpost of the dev that implemented the swift hash ring ( with our consistency tests08:36
nsajeeglynn_: it fails, so I'm off writing a new ML thread for swift :-)08:36
*** flwang1 has joined #openstack-ceilometer08:48
eglynn_nsaje: nice digging! :) ... /me is just catching up with the latest on that ML thread now09:07
*** Longgeek_ has joined #openstack-ceilometer09:13
*** Longgeek has quit IRC09:15
*** ifarkas_ has quit IRC09:15
*** ifarkas has joined #openstack-ceilometer09:17
*** Longgeek_ has quit IRC09:17
*** Longgeek has joined #openstack-ceilometer09:18
*** claudiub has joined #openstack-ceilometer09:25
*** flwang1 has quit IRC09:37
*** Longgeek_ has joined #openstack-ceilometer09:43
DinaBelovansajeб, you have hash ring diving session :)09:46
*** Longgeek has quit IRC09:46
DinaBelovansaje ^^ :)09:47
nsajeyep :)09:47
*** flwang1 has joined #openstack-ceilometer10:00
*** yasukun has quit IRC10:03
openstackgerritA change was merged to openstack/ceilometer: Switch to oslo.serialization
*** nosnos has quit IRC10:14
*** nosnos has joined #openstack-ceilometer10:15
*** nosnos_ has joined #openstack-ceilometer10:18
*** Ugallu has joined #openstack-ceilometer10:18
*** nosnos has quit IRC10:19
*** Ugallu has quit IRC10:22
cmystereglynn_: heya, got some time for me ?10:23
eglynn_cmyster: gimme a minute to finish up something10:23
*** asalkeld has quit IRC10:23
cmystercool, ping us when done10:23
*** flwang1 has quit IRC10:30
*** cdent has joined #openstack-ceilometer10:30
*** k4n0_ has quit IRC10:33
*** k4n0 has joined #openstack-ceilometer10:35
*** _nadya_ has quit IRC10:35
*** flwang1 has joined #openstack-ceilometer10:36
*** Ugallu has joined #openstack-ceilometer10:38
*** samta has quit IRC10:54
eglynn_cmyster: ... so what's up?10:58
cmyster>> pm10:59
*** ajc_ has quit IRC11:25
cdenteglynn_: you about, have a question on:
eglynn_cdent: yeap, what's the question?11:30
cdentAssuming it was only made undone by the WSME + sphinx issues, and assuming those are properly fixed, I think it can probably be declared done again.11:31
*** eglynn__ is now known as eglynn-officeafk11:31
*** arezmerita has joined #openstack-ceilometer11:33
eglynn_cdent: but the "revert" is still in place right?11:33
eglynn_cdent: ... i.e.
cdentAh, hadn't realized that ever went in11:33
eglynn_cdent: wouldn't that have to be un-reverted in order to declare it done again?11:33
cdentOkay, well now would be the wrong time to unrevert that.11:34
cdentWhen we surface from the current gateness, I'll unrevert that and see how things go, shall I?11:34
eglynn_cdent: agree ... prolly one to keep in your pocket until juno-3 is safely in the bag11:34
eglynn_cdent: thanks!11:34
* eglynn_ needs to eat :)11:35
*** eglynn_ is now known as eglynn-lunch11:35
*** flwang1 has quit IRC11:35
* cdent thinks that's a mighty fine idea11:35
*** flwang1 has joined #openstack-ceilometer11:35
openstackgerritZhai, Edwin proposed a change to openstack/ceilometer: Add IPMI support
openstackgerritZhai, Edwin proposed a change to openstack/ceilometer: Add IPMI raw data fetching in IPMI agent
openstackgerritJoe Hakim Rahme proposed a change to openstack/ceilometer: Run unit tests against MySQL
*** underyx has quit IRC11:51
*** underyx has joined #openstack-ceilometer11:51
*** tongli has joined #openstack-ceilometer11:55
*** isviridov is now known as isviridov_away12:08
*** gordc has joined #openstack-ceilometer12:10
*** rbowen has quit IRC12:10
*** rbowen has joined #openstack-ceilometer12:11
*** itisit has joined #openstack-ceilometer12:42
*** itisit has left #openstack-ceilometer12:44
*** eglynn-lunch is now known as eglynn12:58
*** renatoarmani has joined #openstack-ceilometer13:07
openstackgerritgordon chung proposed a change to openstack/python-ceilometerclient: keystone discovery fallback support
DinaBelovagordc, o/13:09
gordcDinaBelova: whatsup13:09
*** itisit has joined #openstack-ceilometer13:10
DinaBelovagordc, about your change
DinaBelovaityaptin is finishing its performance testing right now13:11
DinaBelovawe'll have the results in next 30 mins I guess13:11
*** eglynn-officeafk is now known as eglynn-office13:11
DinaBelovasome approximate result is: your change gives 30-100% improvement13:11
DinaBelovagordc, that is actually really cool :)13:12
gordcDinaBelova: ... really?13:12
*** k4n0 has quit IRC13:12
gordcDinaBelova: that is not expected... i was going to give up on that patch since i didn't see much on my machine.13:12
eglynn-officeDinaBelova, gordc: wow!13:12
eglynn-officeDinaBelova: is that repeatable?13:12
gordcDinaBelova: what db are you using?13:12
DinaBelovathat's ityaptin who's testing it using his scripts right now13:13
ityaptingordc: I'm use mysql13:13
DinaBelovagordc - it's MySQL13:13
gordcityaptin: is it just default settings?13:13
eglynn-officeDinaBelova, ityaptin: 30-100% is a fairly broad range, is that for different types of load scenarios?13:13
DinaBelovaeglynn-office, some minutes are needed13:14
DinaBelovaityaptin has the results, but we need to process them :)13:14
gordcDinaBelova: cool cool... we'll let you guys finish up... but that's pretty good news if that's true.13:14
ityaptineglynn-office, gordc: I will prepare exact results during the next hour and show it.13:14
DinaBelovagordc, np :)13:14
gordcityaptin: thanks!13:14
eglynn-officeityaptin, DinaBelova, gordc: excellent! ... this will be great ammunition in terms of arguing the case for an FFE on the sql improvements13:15
DinaBelovaeglynn-office, yeah, indeed13:15
DinaBelovaeglynn-office, gordc - ityaptin will include lab configuration (the mysql part) as well to the results I guess13:16
gordceglynn-office: agreed. i was going to throw it away when i got 10% increase... but even 30% sounds like a useful patch.13:16
gordcDinaBelova: ityaptin: thanks for all your help.13:17
DinaBelovagordc, pn13:17
* gordc should add you guys as co-authors if patch gets accepted for ffe13:17
DinaBelovagordc :D13:17
eglynn-officegordc: yep, 30% is definitely useful, whereas 100% would be awesome :)13:18
DinaBelovaeglynn-office ;)13:18
cdentDinaBelova: which version of mysql?13:18
nsajeDinaBelova: you guys should do a Ceilometer: before & after (for example Icehouse vs. Juno)13:19
openstackgerritgordon chung proposed a change to openstack/ceilometer: normalise resource data
DinaBelovaityaptin ^^ - may you answer cdent question?13:19
nsajeDinaBelova: so people see the improvements clearly13:19
DinaBelovansaje, we're in process right now :)13:19
nsajeDinaBelova: awesome13:19
DinaBelovansaje, the most nice improvement was the fact that API speed increased :)13:20
*** isviridov_away is now known as isviridov13:20
* gordc wonders if it's worthwhile to just use sqla-core on all the queries... i'll wait for results13:20
cdentWhat's the relative health of gnocchi these days? Did it get as far as hoped?13:27
*** julim has joined #openstack-ceilometer13:27
eglynn-officecdent: jd__ is on vacation this week, we'll discuss at the team meeting next week13:28
*** thomasem has joined #openstack-ceilometer13:38
*** alexpilotti has joined #openstack-ceilometer13:48
*** nosnos_ has quit IRC13:50
*** sbfox has joined #openstack-ceilometer13:50
*** nosnos has joined #openstack-ceilometer13:51
*** r-daneel has joined #openstack-ceilometer13:53
*** nosnos has quit IRC13:55
*** dougwig_ is now known as dougwig14:00
*** rbak has joined #openstack-ceilometer14:08
*** ssakhamuri_ has joined #openstack-ceilometer14:10
*** _nadya_ has joined #openstack-ceilometer14:12
*** underyx_ has joined #openstack-ceilometer14:14
*** underyx has quit IRC14:17
*** Ugallu has quit IRC14:17
*** ekarlso- has quit IRC14:17
*** lsmola has quit IRC14:17
*** harlowja_away has quit IRC14:17
*** ssakhamuri has quit IRC14:17
*** aviau has quit IRC14:17
*** underyx_ is now known as underyx14:17
*** ekarlso- has joined #openstack-ceilometer14:23
*** lsmola has joined #openstack-ceilometer14:25
openstackgerritDina Belova proposed a change to openstack/ceilometer-specs: Add spec about central and compute agents merge.
*** fabiog has joined #openstack-ceilometer14:30
*** lsmola is now known as 18VAAT2X414:33
ityaptineglynn, gordc:
*** Ugallu has joined #openstack-ceilometer14:33
*** 17SAA4BEV has joined #openstack-ceilometer14:33
*** lsmola has joined #openstack-ceilometer14:33
*** ssakhamuri has joined #openstack-ceilometer14:33
*** aviau has joined #openstack-ceilometer14:33
*** 17SAA4BEV has quit IRC14:33
*** lsmola has quit IRC14:34
*** ssakhamuri has quit IRC14:34
*** aviau has quit IRC14:34
ityaptineglynn-office, gordc: Doc with charts and profile results.14:34
*** jaypipes has joined #openstack-ceilometer14:34
eglynnityaptin: excellent :) thank you sir! ... /me reads14:34
*** aviau has joined #openstack-ceilometer14:35
ityaptingordc, cdent: MySQL version 5.5.2814:36
cdentthanks ityaptin14:36
cdentWe've seen pretty big differences between mysql 5.5 and (mysql 5.6 or mariadb 5.5)14:37
*** sbfox has quit IRC14:37
fabiogityaptin: apparently MySQL 5.6 does a better job leveraging indexes than 5.514:37
fabiogityaptin: when you explicitly set an index then 5.5 will perform similarly as 5.6 but in the config we had it wasn't fully leveraging the existing indexes. I believe is related to improvements to the joins14:38
ityaptinfabiog: I can try to update MySQL and restart tests.14:40
gordcityaptin: i believe your current test wouldn't have index i added...14:42
gordcthat said, it'll have index on meter_id, and resource_id individually... just not across both14:42
fabiogityaptin: please try that, you should notice a difference14:42
*** ifarkas has quit IRC14:42
ityaptinfabiog: ok :)14:43
gordcfabiog: would it improve write time? i would think the index is more for read performance?14:43
fabiogityaptin: also you can try to apply only the index that Srini had and see if you notice changes there as we did:
fabioggordc: usually indexes improves read times14:44
cdentgordc: I don't think the issue is with that particular index, but rather with mysql's use of indexes at all14:45
gordccdent: true.14:45
cdentso if there is an upgrade, even without an additional index, there may be differing performance, which is a useful datapoint14:45
fabioggordc: if the db is smart (all are now :-)) it will build the index after the write is completed so it will not degrade writes14:45
fabiogcdent: yes14:45
fabiogcdent: the 5.6 is way more performant than 5.514:45
openstackgerritgordon chung proposed a change to openstack/ceilometer: switch to sqlalchemy core
gordcityaptin: i rebased so it'll have Srinivas' index as well.14:46
fabiogityaptin: we should test the normalize patch and get overall performances on the before and after patch scenario with 5.5 and 5.614:48
ityaptingordc, fabiog, cdent: is any mysql  magic need for test?14:48
fabiogityaptin: but if you want just to see the effect of indexes on 5..5 vs. 5.6 you can add the index that Srini proposed and you will see significant changes (just for your amusement ...)14:48
fabiogityaptin: no, just run meter-list from the client and see the difference in performance (that is what we did)14:49
fabiogor sample-list14:49
gordcfabiog: cdent did something similar already... but with mariadb5.514:49
fabioggordc: yes. I saw that. Thanks cdent that was useful14:50
cdentYou're welcome. It was tedious but I was feeling stubborn.14:50
fabioggordc: please send the pointer to ityaptin14:50
cdentI should have done some 4 color glossies14:50
fabiogcdent: I think we need to apply more due diligence in understanding the improvement in performance we bring in14:51
cdentI agree14:51
*** ildikov has quit IRC14:51
cdentAnd think that in order to have that we need some kind of easy harness for testing that isn't backend dependent14:51
*** Dafna has quit IRC14:52
cdentWe're probably already most of the way there, just need to get everyone on the same page.14:52
fabiogcdent: there is a "general" idea that Ceilometer is not performing well and if we add patches that are not going in the right directions we will give all the wrong signals14:52
* cdent nods14:52
ityaptinfabiog, gordc: ok. I'll try run  tests with 5.5 and 5.6.14:52
fabiogityaptin: super, please keep us posted with your results14:52
ityaptinfabiog, gordc: I'll ping you when it will be complete.14:53
gordcfabiog: agreed. good that we have multiple sources testing these patches.14:53
gordcityaptin: thanks!14:53
fabioggordc: I could not agree more14:53
* cdent wants more hardware14:54
gordcityaptin: also, if possible, could you track how many meters/resources you are creating14:54
gordccdent: same wish here.14:54
ityaptingordc: one moment, please14:55
cdentand with a wave of hands, ityaptin makes new hardware show up on gordc's desk14:55
ityaptinI used 100k of samples and 100 resources14:55
fabiogityaptin: just to make sure that you see what we saw ... you should have a sort of linear degradation when you increase resources14:57
gordcityaptin: cool cool. works for me. just asking because the way db is set up now. it performs better when you have more repeat resources14:57
gordcfabiog: i assume you and i just said the same thing?14:58
ityaptingordc, fabiog: ok) I can define resource count and will run tests with different resources count.14:58
gordccdent: still waiting14:58
gordcityaptin: that's your call. i just wanted to make sure it wasn't a single resource14:59
cdentIt's _invisible_ hardware15:00
gordccdent: funny. i seem to be stocked full of that lately.15:00
cdentMe too. I'd use it but I can never find the power switch.15:01
*** ddieterly has joined #openstack-ceilometer15:02
*** r3pl4y has joined #openstack-ceilometer15:08
openstackgerritLadislav Smola proposed a change to openstack/ceilometer: Adding another set of hardware metrics
*** Longgeek has joined #openstack-ceilometer15:17
*** Longgeek_ has quit IRC15:20
nsajegordc, fabiog : I replied to you guys' comments on
*** rbowen is now known as DrBacchus15:36
*** _cjones_ has joined #openstack-ceilometer15:51
*** packet has joined #openstack-ceilometer15:53
cdentgordc if you want to point me at the client bug I can dig into it this evening, I'm somewhat temporarily at lose ends16:01
gordccdent: i didn't create a bug for it...16:02
gordci think it's related to this:
gordccdent: you can pretty much open any tempest log, and if you look at ceilometer-alarm-evaluator logs... you'll see error16:04
cdentI see if I can replicate it in some local tempest runs16:05
* cdent nods16:05
gordci have this patch:
gordci'm not sure it's related.16:05
cdenti will, as they say, have a play16:05
cdentIs it not odd that the tempest log if full of all that noise but the test passes?16:05
* cdent will take a short walk and then dig16:07
*** sbfox has joined #openstack-ceilometer16:07
*** sbfox has quit IRC16:08
gordccdent: sorry, looked away... yeah... i'm not sure why things are failing... that said, we don't have many tempest tests16:13
*** stevelle_ has joined #openstack-ceilometer16:15
*** r3pl4y has quit IRC16:18
*** _nadya_ has quit IRC16:22
*** stevelle_ has quit IRC16:23
fabiognejc: ping16:28
fabiognsaje: ping16:29
nsajefabiog: o/16:29
fabiognsaje: I have one question regarding the region stuff and the pollsters that I cannot understand16:29
nsajefabiog: fire away16:30
fabiognsaje: if a pollster polls from its region it means that I have to have at least 1 central agent per region, is that right?16:30
nsajeyes, currently it is so16:30
nsajeCeilometer has always polled just stuff in its own region16:30
nsajethat can change in the future, though :-)16:31
fabiognsaje: sure that makes sense16:31
fabiognsaje: the problem is with HA. If that central agent in region A dies, even if the work is partitioned there will be no pollster polling for region A since they are all running in other regions .. do you follow me?16:31
nsajefabiog: yep, I do16:32
nsajefabiog: well, first of all I'd do HA inside each region16:32
*** jergerber has joined #openstack-ceilometer16:32
nsajefabiog: so, multiple central agents running inside each region16:32
nsajefabiog: if all fail, tough luck at the moment, but the discovery mechanism could easily be extended for cross-region endpoint discovery16:33
nsajefabiog: but is that a wise thing to do? It's quite a big amount of traffic to be pumping cross-region16:33
fabiognsaje: yes it is a lot of traffic. But de facto you are defiting the purpose of regions ...16:34
nsajefabiog: I'm not defeating it, since I'm not 'disabling' cross-region functionality, because it doesn't even exist yet :)16:35
nsajefabiog: I'm just not enabling it, yet16:35
fabiognsaje: having multiple regions kind of ensures that if one datacenter is down (really depends on how many AZ you have) you can still survive on the next16:35
nsajefabiog: yes, but if one datacenter is down, why measure it?16:36
fabiognsaje: well it could be only a portion of it that is affected16:36
fabiognsaje: I think adding a sense of "stickiness" to the region is a good thing, but in case there are no pollster from that region others can take over16:37
nsajefabiog: how much do other projects do cross-region? I can't think of a good example, but here goes: if glance is down, does nova in Region1 use glance in Region2 for images?16:37
*** r3pl4y has joined #openstack-ceilometer16:38
fabiognsaje: if you want to reduce or eliminate outage, you probably have to16:38
nsajefabiog: yes, but does it?16:38
fabiognsaje: that is why you have AZ16:39
fabiognsaje to eliminate or reduce this risk16:39
fabiognsaje: now the problem is that in order to do a good job with the central agents you are pushing the burden to the installer/deployer16:40
*** pradk has joined #openstack-ceilometer16:40
fabiognsaje: I can really have HA for Central Agents if I have at least one copy of the Central Agent in every AZ in my region16:41
fabiognsaje: do you agree?16:41
nsajefabiog: well, first I'd like to hear how other projects handle this. I asked about a use-case a few lines back, does nova in Region1 use glance in Region2 if glance in Region1 is down?16:42
fabiognsaje: give me a sec and I will dig it out on how we do cross-region stuff16:45
nsajefabiog: thanks16:46
fabiognsaje: in general though if you have a DNS and a LoadBalancer you can move requests across regions16:46
fabiognsaje: without the service to knwo about it16:46
nsajefabiog: if you for example take a look at this image:
nsajefabiog: I don't think services communicate cross-region16:50
nsajefabiog: among themselves16:50
nsajefabiog: IIUC, AZ are specific to nova, and regions are almost completely separate clouds, apart from Horizon&Keystone16:55
nsajefabiog: so it's operator's responsibility to keep each of the regions healthy, which means running multiple central agents in HA in each region, if necessary16:55
fabiognsaje: the nova guys are taking long to respond ... I will come back to this.16:55
nsajefabiog: but that's just my understanding, I could be completely wrong here. Thanks for taking the time to discuss this!16:56
fabiognsaje: I am trying to make sure that confining the pollsters to regions is not de-facto reducing the HA capabilities16:56
*** eglynn has quit IRC16:57
nsajefabiog: yep, and I completely understand your position. I just want to make sure that we're not over-engineering something that wasn't meant to be in the first place16:57
fabiognsaje: definitely having 1 CA on each AZ of each Region solves it16:57
fabiognsaje: last question and then I will go back to reviewing the patch ....16:58
fabiognsaje: do I need to specify region names/ids in the config of each of these Central Agents or they will be getting this somehow from Keystone?16:59
nsajefabiog: hmm, good question, I guess you'd need to specify it as it is, let me check17:00
*** isviridov is now known as isviridov_away17:01
nsajefabiog: yep, currently it uses the config option, can we get this from keystone?17:01
nsajefabiog: there's a region_name property in keystoneclient.ServiceCatalog, so I guess we can use that17:04
*** chuck_ has joined #openstack-ceilometer17:05
nsajefabiog: I'm not completely sure, but I think keystone returns endpoints in 'our' region only, if no region is passed in17:08
*** sbfox has joined #openstack-ceilometer17:09
*** jbuckley has quit IRC17:15
*** Longgeek_ has joined #openstack-ceilometer17:16
*** Longgeek has quit IRC17:19
*** chuck_ has quit IRC17:19
*** harlowja has joined #openstack-ceilometer17:20
*** claudiub has quit IRC17:28
fabiognsaje: keystone will return the region where the token was issued17:31
nsajefabiog: great, then you can leave the region out in the config17:31
fabiognsaje: considering that in normal cases you will get a token from a "local" keystone we could avoid to have the region in the config17:31
fabiognsaje: this is a good thing to do because it removes the deployment knowledge of the Central Agent17:32
nsajefabiog: I agree17:32
fabiogall the central agents now have the same config17:32
fabiognsaje: please go ahead and make that change and I will review again. thanks a lot. I think we made a significant improvement here17:33
nsajefabiog: I don't know if it's intended or not, but judging by the code, , the default behaviour is to return endpoints in our local region17:36
nsajefabiog: this seems to disable listing all endpoints in all regions together, though17:36
nsajefabiog: do you have somewhere you can test this quickly? otherwise I'll setup something tomorrow (it's almost 8pm here)17:37
fabiognsaje: don't worry I will give it a look. You can however retrieve the region from the token and then make a call to keystone client using the region you got from the token17:41
fabiognsaje: that should keep the current behaviour without having to have the config entry in the config17:42
nsajefabiog: I'll probably do that, I don't think relying on keystoneclient not behaving as intended is wise17:42
nsajefabiog: yep17:42
fabiognsaje: I will put comments in the patch so you can go home now17:42
nsajefabiog: thanks :-)17:42
fabiognsaje: good night, talk to you tomorrow17:43
nsajefabiog: good night!17:43
*** packet has quit IRC17:45
*** packet has joined #openstack-ceilometer17:46
openstackgerritOpenStack Proposal Bot proposed a change to openstack/ceilometer: Updated from global requirements
*** sbfox has quit IRC17:52
openstackgerritOpenStack Proposal Bot proposed a change to openstack/pycadf: Updated from global requirements
*** eglynn has joined #openstack-ceilometer18:01
*** _nadya_ has joined #openstack-ceilometer18:03
*** sbfox has joined #openstack-ceilometer18:14
*** r3pl4y has quit IRC18:17
*** sbfox has quit IRC18:30
*** ddieterly has quit IRC18:31
*** ddieterly has joined #openstack-ceilometer18:32
*** promulo_ has joined #openstack-ceilometer18:34
*** ddieterly has quit IRC18:36
*** nikhil__1 is now known as nikhil18:38
*** nikhil is now known as nikhil_k18:38
*** sbfox has joined #openstack-ceilometer18:38
*** nikhil_k is now known as nikhil__18:40
*** nikhil__ is now known as nikhilkomawar18:42
*** lynxbat has joined #openstack-ceilometer18:44
*** ddieterly has joined #openstack-ceilometer18:46
*** nikhilkomawar is now known as nikhil_k18:48
*** nikhil_k is now known as nikhilk18:49
*** nikhilk is now known as nikhilk_18:50
*** nikhilk_ is now known as nikhil__18:50
*** nikhil__ is now known as nikhil___18:50
*** ddieterly has quit IRC18:51
*** ddieterly has joined #openstack-ceilometer18:51
*** sdake has quit IRC18:51
*** nikhil___ is now known as nikhil_k18:52
*** ddieterly has quit IRC18:55
*** ddieterly has joined #openstack-ceilometer18:56
*** stevelle_ has joined #openstack-ceilometer18:59
gordcpradk: regarding the LBaas, FWaas work.. is there a reason they were implemented the way they were? ie using discovery rather than something like
*** ddieterly has quit IRC19:00
*** ddieterl_ has joined #openstack-ceilometer19:00
pradkgordc, originally thats how i had it in j-1 if you see.. then eglynn suggested to move it to use discovery so we could leverage central agent scaling. lemme dig up his comments19:02
*** itisit1 has joined #openstack-ceilometer19:02
eglynngordc, pradk: the idea was that the central agent partitioning logic was going to be plugged into the discovery extensions19:03
gordcpradk: ah ok... i'll play with it as is...19:03
*** itisit has quit IRC19:03
gordceglynn: cool cool. i need to fix a bug related to it and was wondering if i should make it more like how nsaje implemented discovery items in his current patch.19:04
eglynnyeah the fact that some central agent pollsters still don't use the discovery abstraction turned out to be problematic ... hence
*** ddieterl_ has quit IRC19:04
*** ddieterly has joined #openstack-ceilometer19:05
pradkeglynn, btw thoughts on ? seems like in pipeline.yaml the discovery is scoped more globally and interfereing with generic source19:06
* eglynn looks19:06
pradkeglynn, will nsaje's discovery patch fix this issue as he seems to have moved the sources out of pipeline now?19:06
eglynnhmmm, is that a duplicate of
pradkeglynn, if i negate the meters in generic source all works well.. if not i see dups19:07
eglynnpradk: what do you mean by a "generic source"?19:08
pradkeglynn, the catchall at the top of pipeline.yaml19:08
eglynnpradk: nsaje's patch introduces the concept of per-pollster default discovery, so as to avoid the explosion of sources in the pipeline.yaml19:09
eglynnpradk: ok, this does look like a separate bug to
pradkeglynn, hmm ok .. so the bug you pasted seems related, but the fix you have is just a unit test? at least i still see the issue on mine19:10
pradkeglynn, i discussed briefly with gordc last week and it seemed like the issue was the scope of discovery .. we water to wait until you get back from vacation to see what you thought19:11
eglynnpradk: so the original thought was that the generic catch-all source matching the* was not going to be a problem19:11
eglynnpradk: ... as no resources would be provided to the pollsters in that case19:12
eglynnpradk: ... as the appropriate discovery was not linked to the pollsters in that context19:12
eglynnpradk: ... obviously that reasoning was faulty in some way19:13
eglynnpradk: ... in any case, I'll take that bug and target it to RC119:13
pradkeglynn: cool, i was gonna take a stab at it and see if i can fix the's source to scope the discovery appropriately .. just wanted to wait and see what you thought the issue was19:14
*** _nadya_ has quit IRC19:14
eglynnpradk: a-ha, ok, feel free to grab it back if you want ... otherwise I'll look into it and get a fix into rc119:18
*** packet has quit IRC19:24
*** Longgeek_ has quit IRC19:29
*** lynxbat has left #openstack-ceilometer19:38
openstackgerritFabio Giannetti proposed a change to openstack/ceilometer-specs: Ceilometer RBAC Support for API
*** stevelle_ has quit IRC19:54
*** liusheng has quit IRC20:01
*** liusheng has joined #openstack-ceilometer20:01
*** fabiog has quit IRC20:01
*** liusheng has quit IRC20:02
*** julim has quit IRC20:28
*** flwang1 has quit IRC20:29
cdentgordc I've been poking at the keystone stuff and fell in a deep rabbit hole where I found something that doesn't seem right:
cdentsorry missed a bit:
cdentbased on that, processes started by devstack will have an OS_AUTH_URL with a port of 35357 and path of /v2.0 whereas stuff I do after sourcing openrc will be :5000/v2.020:34
cdentas noted at
cdent35357 does not support /v2.020:35
*** promulo_ has quit IRC20:38
cdentgordc: doing an experiment at:
gordccdent: sounds good. might want to pop over to openstack-keystone if you don't want to wait for gate20:50
*** itisit1 has quit IRC20:50
cdentthanks gordc20:51
*** promulo_ has joined #openstack-ceilometer20:57
*** itisit has joined #openstack-ceilometer20:59
gordccdent: thanks for looking at it. i'll probably be on and off next few days... (brother's getting married)21:05
cdentdo you and eoghan have the same brother?21:05
cdentthe keystone guys say it is a known issue and are looking up a ref21:05
gordccdent: is his brother getting married too?21:08
gordci'm half irish. /joke21:08
nealph_cdent: care to take a look at the PaaS event docco and +1 if appropriate?
gordccdent: ah... awesome21:08
*** sbfox has quit IRC21:09
cdentsweet,  just got +A21:10
cdentbasic bug in keystone, not in "our" stuff21:10
cdentnealph_: would love to, cool if I wait and do it in about 11 hours?21:10
nealph_cdent: could you be more specific on the time?21:11
*** ddieterly has quit IRC21:11
nealph_cdent: I kid, I kid.21:11
cdentgordc: eglyn*'s brother isgetting married yes21:11
cdentit's getting close to my bedtime21:11
nealph_that's totally fine.21:11
*** ddieterly has joined #openstack-ceilometer21:12
* nealph_ thought cdent was US-based...21:12
openstackgerritgordon chung proposed a change to openstack/ceilometer: Migrate the rest of the central agent pollsters to use discoveries
openstackgerritgordon chung proposed a change to openstack/ceilometer: fix network discovery
cdentI'm US-accented, but based in the UK, but also not a morning person...21:12
nealph_ha, fair enough.21:13
*** r3pl4y has joined #openstack-ceilometer21:13
*** ddieterly has quit IRC21:14
*** ddieterly has joined #openstack-ceilometer21:15
*** sbfox has joined #openstack-ceilometer21:15
cdentso I'll make a read of the paas stuff some morning coffee reading21:16
cdentnight all21:16
* cdent waves21:16
*** cdent has quit IRC21:16
*** gordc has quit IRC21:16
*** ddieterl_ has joined #openstack-ceilometer21:17
*** ddieterly has quit IRC21:18
*** thomasem has quit IRC21:23
*** itisit has quit IRC21:24
*** ajc_ has joined #openstack-ceilometer21:29
*** sbfox has quit IRC21:34
*** sbfox has joined #openstack-ceilometer21:38
*** asalkeld has joined #openstack-ceilometer21:38
*** sdake has joined #openstack-ceilometer21:41
*** sdake has quit IRC21:44
*** sdake has joined #openstack-ceilometer21:44
*** sbfox has quit IRC21:47
*** sbfox has joined #openstack-ceilometer21:47
*** tongli has quit IRC21:59
*** asalkeld has quit IRC22:09
*** pradk has quit IRC22:22
*** alexpilotti has quit IRC22:25
*** asalkeld has joined #openstack-ceilometer22:26
*** ddieterl_ has quit IRC22:29
*** ddieterly has joined #openstack-ceilometer22:29
*** sbfox has quit IRC22:31
*** ddieterly has quit IRC22:34
*** sbfox has joined #openstack-ceilometer22:37
*** sbfox has quit IRC22:53
*** rbak has quit IRC23:01
*** r-daneel has quit IRC23:07
*** ajc__ has joined #openstack-ceilometer23:10
*** gzhai has joined #openstack-ceilometer23:10
*** ajc_ has quit IRC23:13
*** eglynn has quit IRC23:20
*** gzhai has quit IRC23:24
*** gzhai has joined #openstack-ceilometer23:31
*** sdake has quit IRC23:57
*** promulo_ has quit IRC23:58

Generated by 2.14.0 by Marius Gedminas - find it at!