Wednesday, 2014-09-03

*** promulo_ has quit IRC00:15
*** promulo_ has joined #openstack-ceilometer00:27
*** nijaba has quit IRC00:38
*** promulo_ has quit IRC00:40
*** r-daneel has quit IRC01:00
*** lnxnut has joined #openstack-ceilometer01:06
*** tongli has quit IRC01:06
*** lnxnut has quit IRC01:16
*** _cjones_ has quit IRC01:19
*** _cjones_ has joined #openstack-ceilometer01:19
*** _cjones_ has quit IRC01:23
*** itisit has joined #openstack-ceilometer01:31
*** asalkeld has quit IRC01:45
*** asalkeld has joined #openstack-ceilometer01:46
*** nosnos has joined #openstack-ceilometer01:52
*** changbl has joined #openstack-ceilometer01:52
*** nosnos has quit IRC02:16
openstackgerritA change was merged to openstack/ceilometer: Imported Translations from Transifex
*** sbfox has quit IRC02:39
*** harlowja is now known as harlowja_away02:52
openstackgerritZhai, Edwin proposed a change to openstack/ceilometer: Add IPMI support
*** cmyster has quit IRC03:21
*** sbfox has joined #openstack-ceilometer03:42
*** ajc_ has joined #openstack-ceilometer04:04
*** changbl has quit IRC04:24
*** sbfox has quit IRC04:31
*** sbfox has joined #openstack-ceilometer04:38
*** itisit has quit IRC04:47
*** nosnos has joined #openstack-ceilometer04:50
openstackgerritRikimaru Honjo proposed a change to openstack/ceilometer: Split bunch of samples into multiple messages
*** yasukun has joined #openstack-ceilometer05:12
*** ildikov has quit IRC05:28
*** sbfox has quit IRC05:37
*** eglynn_ has joined #openstack-ceilometer05:54
openstackgerritOpenStack Proposal Bot proposed a change to openstack/ceilometer: Imported Translations from Transifex
*** k4n0 has joined #openstack-ceilometer06:18
*** _cjones_ has joined #openstack-ceilometer06:27
*** cmyster has joined #openstack-ceilometer06:40
*** _cjones_ has quit IRC06:44
*** _cjones_ has joined #openstack-ceilometer06:45
*** nsaje_ has joined #openstack-ceilometer06:55
*** nsaje has quit IRC06:56
*** ildikov has joined #openstack-ceilometer06:57
*** lsmola has quit IRC07:16
*** nsaje_ has quit IRC07:21
*** nsaje has joined #openstack-ceilometer07:21
*** shardy_afk is now known as shardy07:29
openstackgerritOpenStack Proposal Bot proposed a change to openstack/ceilometer: Updated from global requirements
*** flwang1 has joined #openstack-ceilometer07:37
openstackgerritRikimaru Honjo proposed a change to openstack/ceilometer: Add validate alarm_actions schema in alarm API
*** asalkeld has quit IRC07:38
*** asalkeld has joined #openstack-ceilometer07:39
*** arezmerita has joined #openstack-ceilometer07:42
*** nsaje_ has joined #openstack-ceilometer07:44
*** nsaje has quit IRC07:45
*** nsaje has joined #openstack-ceilometer07:46
openstackgerritA change was merged to openstack/ceilometer: Switch partitioned alarm evaluation to a hash-based approach
*** safchain has joined #openstack-ceilometer07:53
openstackgerritOpenStack Proposal Bot proposed a change to openstack/ceilometer: Updated from global requirements
*** nsaje has quit IRC07:58
*** nsaje has joined #openstack-ceilometer08:01
*** lsmola has joined #openstack-ceilometer08:08
*** krypto has joined #openstack-ceilometer08:11
kryptohi can ceilometer be configured to monitor compute nodes08:11
kryptoas far as i know it can monitor only instance08:12
kryptocan it monitor compute node also?08:12
*** sdake has quit IRC08:16
*** nsaje has quit IRC08:18
*** sdake has joined #openstack-ceilometer08:20
*** jbuckley has joined #openstack-ceilometer08:22
*** drjones has joined #openstack-ceilometer08:24
*** _cjones_ has quit IRC08:28
*** nsaje has joined #openstack-ceilometer08:29
*** drjones has quit IRC08:29
*** Longgeek has joined #openstack-ceilometer08:39
*** nsaje- has joined #openstack-ceilometer08:41
*** nsaje has quit IRC08:41
*** ildikov_ has joined #openstack-ceilometer08:44
*** nsaje- has quit IRC08:45
*** ildikov has quit IRC08:47
*** Longgeek has quit IRC08:53
*** nijaba has joined #openstack-ceilometer08:53
*** nijaba has quit IRC08:53
*** nijaba has joined #openstack-ceilometer08:53
*** Longgeek has joined #openstack-ceilometer08:53
*** foexle has joined #openstack-ceilometer08:59
*** nsaje has joined #openstack-ceilometer09:03
*** nsaje has quit IRC09:08
DinaBelovansaje_, o/09:13
DinaBelovansaje_, do you have any news about that strange discovery behaviour?09:13
*** cdent has joined #openstack-ceilometer09:15
eglynn_DinaBelova: is this a problem with the PoC endpoint discovery patch?09:18
DinaBelovaeglynn_, yeah, indeed09:18
DinaBelovansaje_ and I were debugging it yesterday, I found several issues, nsaje_ told me to do some hacks and there were more issues :)09:19
DinaBelovaeglynn_, I left the comments09:19
eglynn_DinaBelova: cool, /me re-reads the comment trail on gerrit09:19
openstackgerritZhai, Edwin proposed a change to openstack/ceilometer: Add IPMI support
* DinaBelova is debugging and fixing, but I just thought nsaje_ could got some news there09:19
DinaBelovaeglynn_, ok09:19
DinaBelovaif I'll make this change workable on my lab, I'll propose a new patchset I guess09:19
DinaBelovaeglynn_, are you ok with it?09:20
DinaBelovato make the process quicker?09:20
eglynn_DinaBelova: yep, feel free to collaborate with nsaje_ to converge the patch towards correctness09:21
DinaBelovaeglynn_, ok09:21
DinaBelovaeglynn_, I'me testing it in the pair with merged HA change09:21
DinaBelovaso I'll find out if they work ok with each other as well09:21
eglynn_DinaBelova: cool ... in terms of the time horizon, we've already agreed that patch can land post-juno3 as a bugfix, but if we could squeeze it in for tmrw that would be even better09:22
DinaBelovaeglynn_, yeah, I saw the comment in the HA change09:22
DinaBelovaand when ityaptin will be on place, I'll speak with him about gordc change to test09:23
DinaBelova on something small and nice :)09:23
*** Longgeek has quit IRC09:24
eglynn_DinaBelova: excellent :) thank you!09:25
*** Longgeek has joined #openstack-ceilometer09:27
*** nsaje has joined #openstack-ceilometer09:31
eglynn_DinaBelova: just looking at that PoC patch again, should line 89 of ceilometer/ be ... pollster_resources = [, discovery_cache)]09:31
eglynn_DinaBelova: nah, ignore that09:32
DinaBelovaeglynn_, :D09:32
DinaBelovait looks like I found the issue09:32
DinaBelovabut after fix happened nothing :)09:32
ityaptineglynn: Hi! I'll test a gordc's change as soon as possible.09:33
eglynn_DinaBelova: above I meant something more like pollster_resources =[pollster.obj.default_discovery], discovery_cache)09:33
eglynn_ityaptin: cool, thank you sir!09:33
DinaBelovaeglynn_, actually I did the same, ueah09:34
eglynn_DinaBelova: a-ha, but still incorrect behavior?09:34
DinaBelovaeglynn_, yeah09:34
DinaBelovaI guess that discovery extension load went wrong for some reason09:34
* DinaBelova is checking09:34
eglynn_DinaBelova: presumably you've got the 'endpoint = ceilometer.central.discovery:EndpointDiscovery' in your entry_points.txt?09:35
*** nsaje has quit IRC09:36
DinaBelovaeglynn_, I have only these discovery extensions being loaded: ['local_instances', 'lb_pools', 'lb_health_probes', 'fw_services', 'lb_members', 'lb_vips', 'ipsec_connections', 'tripleo_overcloud_nodes', 'fw_policy', 'vpn_services'], endpoint extension load went wrong09:37
DinaBelovaalthough it's set in the setup.cfg09:37
*** nsaje has joined #openstack-ceilometer09:38
eglynn_DinaBelova: have you re-run after applying nsaje's patch, or manually edited the entry_points.txt, to ensure that the setup.cfg change is taking effect?09:39
DinaBelovaeglynn_, /me is checking09:39
DinaBelovaI've rerun the right now09:39
DinaBelovait looks like now after some lines being changed it looks ok09:40
DinaBelovalet's wait polling cycle in one minute :)09:40
eglynn_DinaBelova: cool, so wrapping the pollster.obj.default_discovery in an iterable seems like it does the trick?09:41
DinaBelovaeglynn_, and some lines changes in the discover method + in one place we need to iterate the extensions list, not the ext manager object09:42
DinaBelovaeglynn_, yeah, it looks ok for now09:42
eglynn_DinaBelova: a-ha, cool, good digging! :)09:42
DinaBelovalet me provide some quick fix I guess09:42
eglynn_DinaBelova: if you're push a new version of the patch, can you update the commit message with a link to the bug?09:43
DinaBelovaeglynn_, yeah, sure09:43
eglynn_DinaBelova: thanks!09:44
*** nijaba has quit IRC09:46
*** nijaba has joined #openstack-ceilometer09:46
DinaBelovaeglynn_, btw, about caching lines in the,cm09:48
DinaBelovain the discover method09:48
DinaBelovacurrently it might be list in the discovery param09:49
DinaBelovain this case it'll be error in the dict[list] expression, as list is unhashable09:49
*** nsaje_ has quit IRC09:49
DinaBelovait looks like in this case discover will be list of one item - so we may use discover[0] there, but I'm not sure actually09:50
DinaBelovaeglynn_, what do you think here?09:50
eglynn_DinaBelova: yeah, I see what you mean09:50
cmystereglynn_: heya09:51
eglynn_cmyster: hi09:51
eglynn_DinaBelova: so this method is used for both the pollster-specific and the source-specific discovery09:51
DinaBelovabecause we're wrapping the [pollster.obj.default_discovery] earlier, /me is trying to find out if that'll be the only listing case here >_<09:51
eglynn_DinaBelova: this method being
cmystereglynn_: I am wondering about something here, shouldn't devstack be creating ceilometer:ceilometer ?09:51
DinaBelovaeglynn_, indeed09:52
eglynn_DinaBelova: ... but the logical mismatch is between the source-specific discovery being potentially a list, whereas the pollster-specific is alway just a *single* property09:52
eglynn_cmyster: one sec09:52
DinaBelovaeglynn_, yes, that's ture09:52
eglynn_DinaBelova: ... so how about moving line 183 to inside the for loop, with s/discovery_cache[discovery] = resources/discovery_cache[url] = resources/09:53
eglynn_DinaBelova: ... similarly lines 169-170 would also have to move into the for loop09:54
DinaBelovaa-ha, let me check that09:54
eglynn_DinaBelova: ... for example:09:55
eglynn_for url in (discovery or self.default_discovery):09:55
eglynn_        if discovery_cache and url in discovery_cache:09:55
eglynn_        resources.extend(discovery_cache[url])09:55
eglynn_             continue09:55
eglynn_whoops bad indentation on the line resources.extend(discovery_cache[url]) (needs an extra 4 spaces)09:56
cmysteror -4 in the if line09:56
eglynn_so then the caching is always on the basis of an inidividual discovery URL, not an entire list09:56
eglynn_yeah, wot cmyster says ...09:57
DinaBelovaeglynn_, yeah, I see that, /me is just changing the lines to run some tests one more time to be sure it'll be ok09:57
eglynn_cmyster: you'd a devstack question?09:57
eglynn_DinaBelova: excellent, thank you!09:58
eglynn_cmyster: "ceilometer:ceilometer" is "user:tenant" or?09:58
cmysteras in linux user:group09:58
cmysterpackstack is doing it right, devstack's did not however09:58
eglynn_cmyster: why would devstack need to create these users and groups, I thought it ran everything as sudo?09:59
eglynn_cmyster: ... whereas for packstack, it's installing RPMs and need to do the right WRT to using non-privileged users09:59
cmysterjust creating a group and user should sufice if I need to do the same ?10:00
eglynn_cmyster: if you need to do the same with a devstack, do you mean?10:00
cmysterI'm running wossname's (Kevin?) modified script for 'sanity' testing it and it needs to chown a few things as ceilo:ceilo10:01
eglynn_cmyster: wouldn't you also need to ensure that this user actually owned the ceilometer service processes launched by devstack?10:02
eglynn_cmyster: ... and also that this ceilo user has appropriate perms on /etc/ceilometer, /var/log/ceilometer etc.?10:02
eglynn_cmyster: ... you mean kwhitney?10:02
eglynn_cmyster: ... I thought his sanity script was more oriented to testing an RPM-based install/deploy as opposed to devstack running git clones10:03
cmysteryes him. made his script working for IceHouse but it needs a lot more work10:03
eglynn_cmyster: a-ha, and you want to make it capable of hitting up *both* a devstack and an RPM-based install?10:04
cmystereglynn_: well, its creating data, then creating alerts then running those and checking that the alerts alerted as expected10:04
cmystereglynn_: I think I'll scrap it altogether and rewrite it. it hurts my eyes if you know what I mean10:05
* eglynn_ chuckles and thinks he knows what cmyster means ... ;)10:05
eglynn_cool, that's probably wise TBH10:05
cmysterjust trying to get it to do things in baby steps so I can take those steps and incorporate into something better10:05
*** claudiub has joined #openstack-ceilometer10:06
cmysterthings about babies though, they walk funny ;)10:06
*** nsaje has quit IRC10:06
*** nsaje has joined #openstack-ceilometer10:10
*** nijaba has quit IRC10:12
cdenteglynn: I made some more comments on that IPMI stuff, the most important being the bit about nm_support. I think there must be some pastos going on in that code.10:14
openstackgerritDina Belova proposed a change to openstack/ceilometer: Migrate the rest of the central agent pollsters to use discoveries
cdentThe most recent responses from Edwin raise that common point:10:16
cdente.g. all Fan sensors are reported as same "" without  sensor ID.10:16
cdent(that's a quote)10:16
cdentI believe the same problem came up with disk metrics.10:17
cdentthis is the samples-need-metadata-not-just-resources problem10:18
DinaBelovaeglynn, ^^10:18
DinaBelovaeglynn, I'll check it one more time with multiple central agent running right now10:21
DinaBelovalet's try merge it before tomorrow :)10:22
DinaBelovaeglynn, ehm, I need to understand infra work more I guess... as the patch set was proposed by me was assigned to me as well by OpenStack Infra (hudson-openstack) >_<10:23
*** nsaje has quit IRC10:24
*** asalkeld has quit IRC10:26
*** nsaje has joined #openstack-ceilometer10:27
*** Longgeek has quit IRC10:31
*** Longgeek has joined #openstack-ceilometer10:32
eglynn_cdent: w00t! finally your javelin patch seems to have pleased the gating Gods :) ...
eglynn_cdent: ... that patch should suffice to reinstate the grenade BP onto the juno-3 slate, right?10:34
eglynn_cdent: (... i.e. we're no also gonna block that BP also on the WIP follow-up deeper testing patch?)10:35
eglynn_cool :)10:35
cdentThe belief on the qa side is that the merged patch satisfies the TC's requirements re grenade10:35
cdentwhether that is actually true is unclear10:35
*** Longgeek_ has joined #openstack-ceilometer10:36
eglynn_cdent: well, seeing as Sean is on the TC, I think we can take that interpretation to the bank :)10:37
eglynn_... our TC makes its own truth ;)10:38
* cdent has descended into some kind of weird dystopic wilderness10:38
*** Longgeek has quit IRC10:39
cdentYou weren't kidding when you said (in different words) that grace under compromise is important in this environment10:39
eglynn_"weird dystopic wilderness" ... LOL :)10:42
DinaBelovaeglynn_, I've proposed new patch set, started its testing, found one error by me (will be fixed in next several minutes) and one strange behaviour in the polling if the swift resources by multiple central agents10:46
* DinaBelova is investigating10:46
openstackgerritDina Belova proposed a change to openstack/ceilometer: Migrate the rest of the central agent pollsters to use discoveries
eglynn_DinaBelova: thanks!10:46
*** _nadya_ has joined #openstack-ceilometer10:49
DinaBelovaeglynn_, it looks like it's partitioning the resources by endpoints if it's swift or lbaas, for instance. Like my subset will be ['object_store'], not the metrics10:52
eglynn_DinaBelova: I thought that was the intent of the patch, no?10:56
DinaBelovaeglynn_, so do you think it's okay all swift metrics will be polled by one central agent?10:57
DinaBelovafor instance, in case of the multiple compute agents they share the load via separating list of instances10:57
DinaBelovaand poll everything needed for set of instances10:57
DinaBelovaI supposes central agents should share the load by separating list of containers, for instance here10:58
eglynn_DinaBelova: so the original idea was, IIUC, that the way the glance etc. APIs are constructed, there's not a lot of benefit in spreading that load over multiple central agents10:58
DinaBelovawell, okay, also I have problems with the endpoints :)10:59
DinaBelovaeglynn_, looks like without changing Gordon's comment in the,cm it'll be better11:00
DinaBelovalet me check it11:00
eglynn_DinaBelova: however now that I look at the swift pollster, it seems that we could split per-tenant as opposed to allocating everything to a single central agent11:00
DinaBelovaget_endpoints method is returning urls in the wrong format I guess11:01
DinaBelovaeglynn_, a-ha11:01
eglynn_DinaBelova: nsaje's original version uses service_catalog.get_urls, or?11:01
*** Ephur has quit IRC11:03
* cdent is insufficiently savvy with the data _use_ sides of ceilometer, despite growing savvy with the data gathering11:03
DinaBelovaeglynn_, for the swift changing of get_urls to the get_endpoints makes the world cry actually :)11:03
*** asalkeld has joined #openstack-ceilometer11:03
* eglynn_ assumes that making the world cry is to be avoided at all costs ;)11:04
DinaBelovaeglynn_, yeah, with the nsaje original version it works okay, but one central agent polls all metrics11:04
DinaBelovaprobably I have one container??? :-\11:05
eglynn_DinaBelova: yeah, so I think we could probably do a better job of the splitting for swift, e.g. per tenant11:05
DinaBelovaeglynn_, one moment, it looks like I anyway have one container presented11:05
DinaBelovalet me create some :D11:05
DinaBelovaafter all my dances around this lab only one container survived11:06
eglynn_DinaBelova: ... though a per-container split across agents could be very fine-grained, i.e. potentially a very large number of containers involved11:06
* cdent wants a lab11:06
DinaBelovacdent :D11:06
* cdent and dancing11:06
openstackgerritDina Belova proposed a change to openstack/ceilometer: Migrate the rest of the central agent pollsters to use discoveries
*** Ephur has joined #openstack-ceilometer11:08
DinaBelovaeglynn_, yeah, /me made sure we just have splitting by endpoints and that's it11:09
*** gzhai has joined #openstack-ceilometer11:09
nsajeeglynn_, DinaBelova: what a huge IRC backlog you guys got me11:09
DinaBelovansaje :D11:09
nsajeeglynn_, DinaBelova: I'm gonna need a few minutes to get up to date :-)11:09
eglynn_DinaBelova: cool, let's assume that a per-endpoint split is good enough for now11:10
DinaBelovansaje, sorry, I just made some changes to your change11:10
DinaBelovaeglynn_, yeah11:10
nsajeDinaBelova: np11:10
nsajeDinaBelova: did the change I told you to make yesterday work?11:10
DinaBelovansaje, so now it works ok on multinode devstack lab with HA zookeeper like this: we share the load among multiple central agents by endpoint11:11
DinaBelovansaje, yeah11:11
DinaBelovawith some more lines changed :)11:11
nsajeDinaBelova: that's great!11:11
DinaBelovasome moments with the caching and some moments with the things we are iterating - like not the ext manager, but actually the extensions inside, etc.11:12
DinaBelovansaje, although it looks like we might have some improvements in the follow-up patches11:12
nsajeDinaBelova: great11:12
DinaBelovalike more intelligent sharing the load among central agents11:12
DinaBelovansaje - like eglynn_ proposed, share the load by tenants, not by only endpoints11:13
DinaBelovansaje, because for the swift, for instance, it makes much sence11:13
nsajeDinaBelova: that's only possible for some pollsters, like swift (that's in the works)11:13
nsajeDinaBelova: ah, yeah, I just didn't want to put it in one patch11:13
DinaBelovansaje, ok, cool11:13
DinaBelovaare you working on the swift pollster right now?11:14
nsajeDinaBelova: nope, need to fix up the tests for the current patch first11:14
nsajeDinaBelova: quite some changes, so it's a bit of a chore11:14
DinaBelovansaje, ok, cool, thanks so much!11:14
DinaBelovansaje, what kind of estimations do you propose for this work?11:14
eglynn_nsaje: sounds wise, splitting further improvements over several patches is a god idea (... especially we still harbour a faint hope of squeezing the main patch into juno-3)11:15
nsajeDinaBelova: tests up this evening11:15
DinaBelovaI just think it'll be cool to merge this change before milestone cut11:15
DinaBelovayahoo :)11:15
eglynn_nsaje: cool :)11:15
DinaBelovansaje, it's actually really nice11:15
cdentThis whole hashring distributed coordination whooha seems like a nice feather in the ceilo cap. Makes the project seem a bit cooler and smarter than before.11:17
DinaBelovacdent, yeah, that's actually true :)11:17
nsajeLet's just hope it doesn't fall apart in production :-) *fingers crossed*11:17
cdentIf does nsaje we'll all blame you and you won't get any presents!11:18
DinaBelovafolks, don't worry :)11:18
DinaBelovawe'll fix it quicker it'll be found in production I guess :D11:18
* DinaBelova hopes11:18
* cdent puts nsaje on his xmas list11:18
nsajebetter check it twice ;)11:19
* DinaBelova thinks cdent is Santa and he did not bring Dina any presents last time11:19
cdentlast time cdent was not aware of DinaBelova's existence, apparently a fault in the list software11:20
*** krypto has quit IRC11:20
*** deepthi has joined #openstack-ceilometer11:20
nsajeperhaps the list software uses hash ring partitioning and it didn't work yet11:21
cdentAre you registered with ChristmasKeystone DinaBelova ?11:22
* DinaBelova is wondering what is that11:22
DinaBelovaheh, not yet :(11:23
cdentThat explains it then.11:26
* cdent needs lunch11:26
*** cdent has quit IRC11:26
eglynn_nsaje: I'm a bit confused by the ML responses to your hash-ring thread
nsajeeglynn_: yeah, I need to reply11:27
eglynn_nsaje: IIUC the problem with the Ironic approach was that adding a new *conductor* causes a unlimted number of nodes to be re-assigned ... is that correct?11:28
nsajeeglynn_: yes. I'll need to state the problem more specifically11:28
DinaBelovajd__, sorry for disturbing, but /me goes crazy with - do you have any idea what is that strange thing??? it looks like I don't see something really easy and small...11:28
nsajeeglynn_: they don't understand their implementation yet :-)11:28
DinaBelovansaje lol11:29
eglynn_nsaje: cool :) ... so maybe slightly clearer as s/when you add a new host/when you add a new conductor/11:29
eglynn_(... as folks may be thinking host == baremetal-node)11:29
nsajeeglynn_: yeah, but they have conductors as 'hosts' in the hash ring code11:29
nsajeeglynn_: that's why I put it in there :)11:29
eglynn_nsaje: a-ha, got it :)11:30
eglynn_nsaje: similarly 1/#nodes means 1/#conductors, or?11:30
nsajenope, that actually means baremetal nodes11:30
eglynn_nsaje: a-ha, k, cool11:30
nsajeeglynn_: just a sec, let me check. It's starting to confuse me as well, lol11:31
nsajeeglynn_: sorry, yes, in the e-mail, 1/#nodes means 1/#conductors11:31
nsajeeglynn_: I'll make that clearer, thank you11:31
eglynn_nsaje: coolness :)11:31
nsajeeglynn_: replied to both11:50
eglynn_nsaje: thanks! ... /me looks11:50
nsajeeglynn_: oops, missed a reference in the 2nd email11:52
*** Longgeek_ has quit IRC11:53
*** Longgeek has joined #openstack-ceilometer11:54
*** Longgeek_ has joined #openstack-ceilometer11:54
nsajeeglynn_: it should be [1]
*** deepthi has quit IRC11:57
*** Longgeek has quit IRC11:58
*** lsmola has quit IRC12:01
nsajeDinaBelova: can you take a look at, L213 please?12:02
* DinaBelova looking12:02
nsajeDinaBelova: is there a point in creating a discovery_cache if it doesn't exist?12:02
nsajeDinaBelova: If it's not passed into the function, what use is there for it?12:03
DinaBelovansaje, well, because I supposed we need to put the partitioned to the discovery cache12:03
DinaBelovaas it's dictionary that is passed time by time12:04
DinaBelovait'll be travelling from polling cycle to polling cycle12:04
nsajeDinaBelova: yes, but shouldn't it be created outside the function? otherwise it won't survive between function executions12:04
DinaBelovahm, one moment12:04
DinaBelovainteresting question12:05
nsajeDinaBelova: it's already created in L8912:05
DinaBelovaI see that, but also I had errors with None cache there12:05
nsajeDinaBelova: we should probably just have12:06
DinaBelovansaje, probably that's because this discovery method is pollster-specific and the source-specific discovery12:06
nsajeif discovery_cache:12:06
nsaje    discovery_cache[url] = ...12:06
DinaBelovawell, I got your idea, but what will be do with the discovered and partitioned info?12:07
DinaBelovaif it'll be None12:07
nsajethen there's no caching :-) it's the caller's responsibility to pass in a cache12:07
DinaBelovawell, yeah, it looks like I was a little bit crazy here12:07
DinaBelovaplease feel free to remove this code, nsaje  :)12:08
nsajeDinaBelova: cool!12:08
DinaBelovansaje, it looks like some over-work thing happened in my mind :D12:08
nsajeDinaBelova: hehe :-)12:08
*** lsmola has joined #openstack-ceilometer12:10
eglynn_nsaje: cool, I'll throw in my $0.02 on the thread also12:10
nsajeDinaBelova: I have some branch divergence going on right now, please tell me, you only modified the file after my last patch?12:13
DinaBelovansaje, yeah, and removed one line in lbaas module12:14
DinaBelovaactually Itryed to fix all comments you had on your last patch set12:15
DinaBelovaGordon's was actually the thing I should not change12:15
DinaBelovaafter some checking it turned out it was not worth doing12:15
DinaBelovansaje, so and lbaas.py12:15
*** nosnos has quit IRC12:16
openstackgerritZhai, Edwin proposed a change to openstack/ceilometer: Add IPMI support
*** nosnos has joined #openstack-ceilometer12:17
nsajeDinaBelova: thanks12:17
*** Longgeek_ has quit IRC12:19
openstackgerritIlya Tyaptin proposed a change to openstack/ceilometer: [HBase] Catch AlreadyExists error in Connection upgrade
*** nosnos has quit IRC12:21
nsajeeglynn_: thanks for helping out on the ML, you make explaining something look so easy, even though it's anything but :-)12:25
eglynn_nsaje: np! ... hope it helped clarify12:26
openstackgerritRikimaru Honjo proposed a change to openstack/ceilometer: Add validate alarm_actions schema in alarm API
*** Longgeek has joined #openstack-ceilometer12:30
*** asalkeld_ has joined #openstack-ceilometer12:31
openstackgerritA change was merged to openstack/pycadf: define the project goal
*** asalkeld has quit IRC12:33
*** cdent has joined #openstack-ceilometer12:36
*** asalkeld_ is now known as asalkeld12:37
*** asalkeld has left #openstack-ceilometer12:38
*** yasukun has quit IRC12:38
*** itisit has joined #openstack-ceilometer12:40
cdenteglynn: This IPMI thing is getting challenging to track.12:40
DinaBelovaeglynn, may you take a look on ? It's +2'ed by Nadya ~2 weeks go and all other core team is afraid to touch HBase change :)12:48
eglynn_DinaBelova: the rel mgr has just requested that there are no further patch approvals before juno-3 to minimize load on the gate12:49
openstackgerritVitaly Gridnev proposed a change to openstack/ceilometer: Update docs with Sahara notifications configuration
DinaBelovaeglynn_ :(12:50
DinaBelovathat's sad12:50
eglynn_jd__, sileht, ildikov_, _nadya_, llu: the rel mgr has just requested that there are no further patch approvals before juno-3 to minimize load on the gate12:50
cdentrel mgr is drunk with power! ;)12:50
eglynn_DinaBelova: well, it's not that big an issue, as all the bugs will just get bumped to juno-rc112:51
eglynn_DinaBelova: ... so that patch hasn't missed juno12:51
DinaBelovaeglynn_, yeah, I just think I need to go through old changes more often :D12:52
*** nijaba has joined #openstack-ceilometer12:52
*** nijaba has quit IRC12:52
*** nijaba has joined #openstack-ceilometer12:52
*** igor has joined #openstack-ceilometer12:52
*** ajc_ has quit IRC12:52
eglynn_cdent: yeah, do my comments about mixing the per-sensor and system-wide data in the same meter make sense to you?12:52
*** igor has quit IRC12:52
*** igor has joined #openstack-ceilometer12:53
silehteglynn_, I have seen that, gate merge queue is 17 hr 47 min long12:53
eglynn_sileht: yeah, does that qualify as record? ... sadly, probably not12:53
cdenteglynn_: yes, they make sense and seem to make some kind of sense to Edwin as well as he's proposed a solution12:54
cdentBut the twisty turny passages are making it hard to be clear on if its right12:54
silehteglynn_, I think we have made a better score in the past, with some sporadic issues gate :p12:55
*** igor has quit IRC12:56
* cdent fetches car and goes home back in about 30 minutes12:56
*** cdent has quit IRC12:56
_nadya_eglynn_: ack12:57
*** changbl has joined #openstack-ceilometer13:00
*** eglynn_ is now known as eglynn-lunch13:07
*** _nadya_ has quit IRC13:10
openstackgerritRich Bowen proposed a change to openstack/ceilometer: Adds some details and examples for alarming
openstackgerritRich Bowen proposed a change to openstack/ceilometer: Adds some details and examples for alarming
*** tongli has joined #openstack-ceilometer13:20
openstackgerritRich Bowen proposed a change to openstack/ceilometer: Adds some details and examples for alarming
*** gordc has joined #openstack-ceilometer13:21
*** julim has joined #openstack-ceilometer13:24
ildikov_eglynn: ok, thanks for the info13:26
*** thomasem has joined #openstack-ceilometer13:28
ildikov_eglynn: the docco is merged! \o/ :)13:29
*** thomasem has quit IRC13:34
*** thomasem has joined #openstack-ceilometer13:35
openstackgerritRikimaru Honjo proposed a change to openstack/ceilometer: Split bunch of samples into multiple messages
*** flwang1 has quit IRC13:39
*** jasondotstar has joined #openstack-ceilometer13:39
*** r-daneel has joined #openstack-ceilometer13:47
openstackgerritNejc Saje proposed a change to openstack/ceilometer: Migrate the rest of the central agent pollsters to use discoveries
*** cdent has joined #openstack-ceilometer13:56
*** dhellmann has quit IRC13:57
*** dhellmann has joined #openstack-ceilometer13:58
*** nsaje_ has joined #openstack-ceilometer14:00
*** nsaje is now known as nsaje-office14:00
*** nsaje_ is now known as nsaje14:00
*** gzhai has quit IRC14:04
*** rbak has joined #openstack-ceilometer14:09
cdenteglynn: I'm home now. What's the state of the IPMI stuff, still needing a poke?14:16
eglynn-lunchcdent: would be great if you could verify the new agent in the non-Intel case14:28
eglynn-lunchcdent: ... I gotta step out briefly now, back in a few mins14:28
*** eglynn-lunch is now known as eglynn-afk14:28
cdentI'll try, but since the endpoint is a stub, not sure if it will work.14:29
*** ildikov has joined #openstack-ceilometer14:36
*** ildikov_ has quit IRC14:40
*** ildikov has quit IRC14:41
*** shengjiemin has quit IRC14:51
*** arezmerita has quit IRC14:56
*** sbfox has joined #openstack-ceilometer14:57
*** arezmerita has joined #openstack-ceilometer15:02
*** _nadya_ has joined #openstack-ceilometer15:08
*** eglynn-afk is now known as eglynn_15:13
cdenteglynn_: I've done some digging15:21
eglynn_cdent: cool, do share :)15:22
cdentThe latest patchset plus my testing with lucas' ipmi host suggests that when the code address a host that does not support node manager it will give artificial stub data15:22
eglynn_cdent: "artificial" == "invented data"?15:23
cdentlook at read_power_all here:,cm15:23
* eglynn_ looks15:24
cdentnm_support will only be true if a raw request via ipmitool returns the right bits (literal) that indicate node manager support15:24
cdentI'm currently trying to determine if that's a commo thing or not15:25
cdentbut the code isn't making a ton of sense15:25
cdent(because of the indirection of the decorator)15:26
eglynn_cdent: yeah good spot, the "Get dummy data on platfrom without Node Manager" is sending shivers down my spine15:26
cdentMy guess is that the response will be "we plan to do that later"15:29
eglynn_cdent: yeah probably, though that was not my understanding of Edwin's reponse to my questions on the review about what happens on non-Intel platforms15:30
eglynn_... maybe I misunderstood, or wasn't specific enough in my question15:30
eglynn_cdent: anyhoo, good digging, can you make a note on the review of your findings/suspicions15:31
cdentYeah, will do15:31
*** alexpilotti has joined #openstack-ceilometer15:34
*** _cjones_ has joined #openstack-ceilometer15:43
*** _cjones_ has quit IRC15:49
*** _cjones_ has joined #openstack-ceilometer15:49
*** arezmerita has quit IRC15:50
cdentcomments up eglynn_15:54
eglynn_cdent: thank you sir! ... looking now15:55
*** foexle has quit IRC15:56
*** arezmerita has joined #openstack-ceilometer15:57
*** Longgeek has quit IRC16:02
*** eglynn__ has joined #openstack-ceilometer16:04
cdentgordc: you have any great learnings on the database front?16:04
*** mrutkows has joined #openstack-ceilometer16:05
*** drjones has joined #openstack-ceilometer16:08
*** ondergetekende_ has joined #openstack-ceilometer16:08
*** xmltok_ has joined #openstack-ceilometer16:08
*** Ugallu has joined #openstack-ceilometer16:08
*** theanalyst has quit IRC16:08
*** ondergetekende has quit IRC16:08
*** jasondotstar has quit IRC16:08
*** k4n0 has quit IRC16:08
*** sileht has quit IRC16:08
*** eglynn has quit IRC16:08
*** dougwig has quit IRC16:08
*** _cjones_ has quit IRC16:08
*** arezmerita has quit IRC16:08
*** xmltok has quit IRC16:08
gordccdent: sort of... i mean batching is possible... but not without caching i don't think... or if not caching, something even more heavyweight(ie.taskflow/tooz type stuff)16:08
*** theanalyst has joined #openstack-ceilometer16:09
*** k4n0_ has joined #openstack-ceilometer16:09
*** arezmerita has joined #openstack-ceilometer16:09
*** dougwig_ has joined #openstack-ceilometer16:10
*** sileht has joined #openstack-ceilometer16:11
cdentI suspect a two stage write might  be workable: immeidately write raw data to some highly available store and then use that as a queue for writing to a more structure store.16:11
cdentBut that's a big change.16:11
gordccdent: yeah, that's too big.16:20
cdentyeah, not suggesting it as a now by any stretch16:20
gordccdent: i'm not sure i can think of a way that can be contained within juno16:21
cdentyeah based on zzzeek's comments the only fixes are big changes16:21
gordccdent: yeah. theh caching work isn't big... but i don't think it offers much value since profiling showed it was transaction16:22
gordcnumber which was slowing us down16:23
* cdent nods16:23
gordc11ms per transaction does seem a bit high though.16:28
*** safchain has quit IRC16:28
cdentgordc: have you recorded the actual SQL that gets generated during a worst case of one of those transactions?16:29
gordcnope. i havent actually looked at it... i assume worse case you mean, it creates meter, resource, and sample16:30
cdentI _think_ so :)16:31
cdentI guess, however, that Mike's point, though was that the cost is estabishing the transaction, not actually running the queries?16:32
gordcyeah. i thikn it's the actual commit that is slow.16:32
gordcwhich seems strange... i'll dig into that. i think i might have another idea but i'll hash it out during lunch.16:33
* cdent waves16:33
*** _nadya_ has quit IRC16:36
*** amalagon has joined #openstack-ceilometer16:39
*** ildikov has joined #openstack-ceilometer16:41
*** claudiub has quit IRC16:51
openstackgerritA change was merged to openstack/ceilometer: XenAPI support: Disk rates
*** dougwig_ is now known as dougwig17:00
*** palar has joined #openstack-ceilometer17:11
*** renatoarmani has joined #openstack-ceilometer17:18
*** harlowja_away is now known as harlowja17:20
*** jbuckley has quit IRC17:21
*** mrutkows has quit IRC17:32
*** alexpilotti has quit IRC17:39
*** alexpilotti has joined #openstack-ceilometer17:41
*** eglynn_ has quit IRC17:51
*** eglynn_ has joined #openstack-ceilometer17:51
*** palar_ has joined #openstack-ceilometer18:12
*** palar has quit IRC18:15
*** mspreitz has joined #openstack-ceilometer18:17
openstackgerritDoug Hellmann proposed a change to openstack/ceilometer: warn against sorting requirements
*** palar_ has quit IRC18:33
*** _nadya_ has joined #openstack-ceilometer18:41
*** Ugallu has quit IRC18:45
*** _nadya_ has quit IRC18:46
*** promulo_ has joined #openstack-ceilometer18:50
*** claudiub has joined #openstack-ceilometer18:56
mspreitzIt looks to me like python-ceilometerclient has been broken since August 7.
gordcmspreitz: are you taking a look at that bug?19:25
mspreitzI just diagnosed it.  I was hoping someone more expert would prepare the fix19:25
gordcmspreitz: ok cool. i'll try having a look. i need a break from current business19:26
mspreitzgordc: thanks!19:26
*** _nadya_ has joined #openstack-ceilometer19:29
*** renatoar_ has joined #openstack-ceilometer19:29
*** renatoarmani has quit IRC19:29
*** _nadya_ has quit IRC19:33
*** _nadya_ has joined #openstack-ceilometer19:34
*** jergerber has joined #openstack-ceilometer19:35
*** _nadya_ has quit IRC19:39
*** itisit has quit IRC19:42
*** lynxbat has joined #openstack-ceilometer19:46
*** amalagon has quit IRC19:46
*** flwang1 has joined #openstack-ceilometer20:13
*** itisit has joined #openstack-ceilometer20:13
*** flwang1 has quit IRC20:21
openstackgerritgordon chung proposed a change to openstack/python-ceilometerclient: sync oslo code
*** julim has quit IRC20:24
*** gordc has quit IRC20:24
openstackgerritJeremy Stanley proposed a change to openstack/ceilometer: Work toward Python 3.4 support and testing
openstackgerritJeremy Stanley proposed a change to openstack/pycadf: Work toward Python 3.4 support and testing
openstackgerritJeremy Stanley proposed a change to stackforge/gnocchi: Work toward Python 3.4 support and testing
*** cdent has quit IRC20:32
*** ildikov has quit IRC20:43
mspreitzgordc: any progress on bug 1364760 ?20:54
*** lynxbat has quit IRC21:11
*** asalkeld has joined #openstack-ceilometer21:17
*** thomasem has quit IRC21:22
*** dhellmann is now known as dhellmann_21:28
*** dekozo has quit IRC21:33
*** eglynn_ has quit IRC21:36
*** nsaje has quit IRC21:37
*** lynxbat has joined #openstack-ceilometer21:46
*** lynxbat has quit IRC21:46
*** lynxbat has joined #openstack-ceilometer21:47
*** asalkeld has quit IRC21:47
*** lynxbat has quit IRC22:00
*** lynxbat has joined #openstack-ceilometer22:03
*** asalkeld has joined #openstack-ceilometer22:03
*** shardy is now known as shardy_z22:07
*** openstack has joined #openstack-ceilometer22:10
*** asalkeld has quit IRC22:13
*** sdake_ has joined #openstack-ceilometer22:18
*** packet has joined #openstack-ceilometer22:22
*** asalkeld has joined #openstack-ceilometer22:30
*** asalkeld_ has joined #openstack-ceilometer22:37
*** sdake_ has quit IRC22:37
*** asalkeld has quit IRC22:39
*** asalkeld_ has quit IRC22:39
*** yasukun has joined #openstack-ceilometer22:46
*** asalkeld has joined #openstack-ceilometer22:46
*** promulo_ has quit IRC22:49
*** itisit has quit IRC22:50
*** promulo_ has joined #openstack-ceilometer22:54
openstackgerritPhil Neal proposed a change to openstack/ceilometer: Document the standard for PaaS service notifications
*** rbak has quit IRC22:58
*** jergerber has quit IRC23:06
*** promulo_ has quit IRC23:15
*** asalkeld has quit IRC23:15
*** openstackstatus has quit IRC23:19
*** openstackstatus has joined #openstack-ceilometer23:19
*** ChanServ sets mode: +v openstackstatus23:19
*** promulo_ has joined #openstack-ceilometer23:22
*** asalkeld has joined #openstack-ceilometer23:32
*** alexpilotti has quit IRC23:37
*** promulo_ has quit IRC23:37
*** renatoar_ has quit IRC23:45

Generated by 2.14.0 by Marius Gedminas - find it at!