Friday, 2015-08-07

*** ankita_w_ has joined #openstack-ceilometer00:44
*** ankita_wagh has quit IRC00:44
*** liusheng has joined #openstack-ceilometer00:55
*** thorst has joined #openstack-ceilometer01:16
*** thorst has quit IRC01:17
*** ankita_w_ has quit IRC01:18
*** tongli has joined #openstack-ceilometer01:18
*** pauldix has joined #openstack-ceilometer01:34
*** pauldix has quit IRC01:40
*** tongli has quit IRC02:16
*** _cjones_ has joined #openstack-ceilometer02:36
*** _cjones_ has quit IRC02:37
*** _cjones_ has joined #openstack-ceilometer02:37
*** kun_huang has joined #openstack-ceilometer02:49
openstackgerritliusheng proposed openstack/aodh: WIP: Use storage scenario test base to test migration
*** rjaiswal has quit IRC03:40
*** _cjones_ has quit IRC03:40
*** boris-42 has quit IRC03:50
*** ankita_wagh has joined #openstack-ceilometer03:53
*** r-mibu has joined #openstack-ceilometer03:55
*** dikonoor has joined #openstack-ceilometer04:06
*** ankita_wagh has quit IRC04:09
*** ankita_wagh has joined #openstack-ceilometer04:09
*** ildikov has quit IRC04:29
*** jruano has joined #openstack-ceilometer04:39
*** jdandrea has quit IRC04:39
*** jruano has quit IRC04:58
*** ankita_w_ has joined #openstack-ceilometer05:10
*** ildikov has joined #openstack-ceilometer05:12
*** ankita_wagh has quit IRC05:13
*** yprokule has joined #openstack-ceilometer05:16
*** belmoreira has quit IRC05:47
*** nadya has joined #openstack-ceilometer06:27
*** belmoreira has joined #openstack-ceilometer06:38
openstackgerritMehdi Abaakouk (sileht) proposed openstack/gnocchi: Implements list_resources limit/ordering
*** idegtiarov has joined #openstack-ceilometer07:13
*** pauldix has joined #openstack-ceilometer07:37
*** ankita_w_ has quit IRC07:41
*** pauldix has quit IRC07:41
openstackgerritliusheng proposed openstack/aodh: Use storage scenario test base to test migration
*** safchain has joined #openstack-ceilometer07:58
*** liusheng has quit IRC08:00
openstackgerritMehdi Abaakouk (sileht) proposed openstack/gnocchi: rest: Add links to /v1 endpoint
openstackgerritMehdi Abaakouk (sileht) proposed openstack/gnocchi: rest: return metrics for ../<resource_id>/metric
openstackgerritMehdi Abaakouk (sileht) proposed openstack/gnocchi: rest: fix archive policy controller
*** krypto has joined #openstack-ceilometer08:07
*** alexpilotti has joined #openstack-ceilometer08:07
kryptohi for restarting a failed instance using heat,does it need trigger from ceilometer08:08
*** yassine_ has joined #openstack-ceilometer08:15
*** afazekas_ has joined #openstack-ceilometer08:17
zigoI got some unit test errors in MySQL when building tooz:
*** kun_huang has left #openstack-ceilometer08:21
zigoThis could be because I'm using the last version of MySQL in Debian (ie: 5.6).08:22
zigoAlso, the is very naive, the "wait_for_line" simply doesn't work.08:22
zigoI replaced this by a mysqladmin ping, this works a way better.08:22
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
zigoOk, I figured it out.08:32
zigoNow I have other issues.08:33
zigosileht: Any idea? ^08:33
* sileht reads08:34
zigosileht: WTF this piece of code is attempting? :)08:39
openstackgerritJulien Danjou proposed openstack/aodh: devstack: only use 1 thread
silehtzigo, I don't see how the file can be empty08:44
zigosileht: What can I do to help you to help?08:46
silehtzigo, do you have the same issue on py27, or this is a python3 issue ?08:48
zigosileht: It's looking like a Py3 only problem.08:49
zigoNo unit test failures under Py2.708:50
zigoMaybe concurrent.futures is different in Py3 ?08:50
silehtyes it's different08:51
openstackgerritIgor Degtiarov proposed openstack/ceilometer: Cast Int64 values to float
zigosileht: Also, maybe I have a different version as the Trusty one, as Sid got updated but not Trusty.08:51
zigosileht: What I can do is upload tooz 1.21.0 with disabled Py3 tests, then you can try to build it re-enabling it.08:52
zigosileht: Does this work?08:52
silehtzigo, I'm building it locally :)08:53
zigosileht: In Sid ?08:53
silehtin jessie to see I got the issue, because py3 jessie != sid08:53
zigoI've uploaded to Experimental with disabled Py3 tests.08:57
zigoThe package is at ssh:// in the debian/liberty branch.08:57
*** dikonoor has quit IRC08:58
zigoI filed this:
openstackLaunchpad bug 1482555 in tooz "Py3 unit test failures when building debian package" [Undecided,New]09:05
*** dikonoo has joined #openstack-ceilometer09:07
silehtzigo, you have a bug in the script that run unittest (unrelated with the tooz bug you have)09:10
zigosileht: What bug?09:10
silehtzigo, py3 tests doesn't run by default and the building doesn't fail09:10
zigosileht: I just told you, I disabled Py3 tests so I could upload.09:11
silehtlol sorry :p09:11
zigosileht: In the line set -e ; set -x ; for i in 2.7 ; do09:11
zigoJust add $(PYTHON3S) after 2.709:11
silehtarf I can't build it in jessie my pbr is too old09:12
zigosileht: Yeah, you should try building in Sid + Experimental.09:16
zigoI haven't worked-out the Jessie backport repos yet.09:17
zigoThese days, I've been working mostly on plain Sid directly.09:17
zigoAnwyay, time to go back home now, and prepare the trip for Debconf 15 ! :)09:17
*** krypto has quit IRC09:29
openstackgerritMehdi Abaakouk (sileht) proposed openstack/gnocchi: rest: return metrics for ../<resource_id>/metric
openstackgerritMehdi Abaakouk (sileht) proposed openstack/gnocchi: rest: fix archive policy controller
openstackgerritMehdi Abaakouk (sileht) proposed openstack/gnocchi: rest: Add links to /v1 endpoint
openstackgerritJulien Danjou proposed openstack/aodh: storage: remove deprecated database_connection
openstackgerritJulien Danjou proposed openstack/aodh: storage: re-add and deprecate alarm_connection
*** tobias- has quit IRC10:17
*** tobias- has joined #openstack-ceilometer10:17
*** divya__ has joined #openstack-ceilometer10:19
*** dikonoo has quit IRC10:22
*** cdent has joined #openstack-ceilometer10:44
*** safchain has quit IRC10:45
*** harlowja has quit IRC10:45
*** openstackgerrit has quit IRC10:45
*** tobasco has quit IRC10:45
*** _gryf has quit IRC10:45
*** aignatov has quit IRC10:45
*** raginbajin has quit IRC10:45
*** dhellmann has quit IRC10:45
*** d0ugal has quit IRC10:45
*** therve has quit IRC10:45
cdentthanks for the on aodh upgrade stuff jd__, your answer is pretty much what I was expecting (including the need for an alarm_connection fixup), but I wanted to get it explicit and out under the eyes of anyone interested. IRC is such a horrible place to figure out anything with any real understanding10:46
*** safchain has joined #openstack-ceilometer10:49
*** harlowja has joined #openstack-ceilometer10:49
*** openstackgerrit has joined #openstack-ceilometer10:49
*** _gryf has joined #openstack-ceilometer10:49
*** tobasco has joined #openstack-ceilometer10:49
*** aignatov has joined #openstack-ceilometer10:49
*** raginbajin has joined #openstack-ceilometer10:49
*** dhellmann has joined #openstack-ceilometer10:49
*** d0ugal has joined #openstack-ceilometer10:49
*** therve has joined #openstack-ceilometer10:49
openstackgerritJulien Danjou proposed openstack/aodh: storage: re-add and deprecate alarm_connection
*** yassine_ has quit IRC11:28
*** gordc has joined #openstack-ceilometer11:35
openstackgerritMerged openstack/aodh: Use storage scenario test base to test migration
*** thorst has joined #openstack-ceilometer11:40
openstackgerritgordon chung proposed openstack/ceilometer: disable non-metric meter definitions
nadyagordc: Gordon, hi! Am I right, that one message cannot be processed by two pipelines: event's and sample's? The message is taken from mq by one of them?11:58
gordcnadya: not quite.12:00
gordca message will be picked up by a listener and then it will be dispatched to multiple endpoints, some endpoints will convert to sample, some will convert to events.12:01
gordconce it's either a sample or an event, it will be processed by the appopriate pipeline.12:01
gordc(as long as there is no bug in the code)12:02
nadyagordc: aha, yep, I see the code...12:02
nadyagordc: and maybe there is a bug there :)12:02
gordc:| i hope not... there's a test... unless there's also a bug in the tests...12:02
nadyagordc: will continue investigations, because I see meters but don't see corresponding events12:03
gordctests for tests?12:03
gordcoh. are you on master?12:03
gordcdo you have store_events enabled?12:03
nadyagordc: good news that on devstack with Ceilo master everything is fine (I hope that info is correct)12:04
nadyagordc: yep, store_events is on. I'll try to find related bugs, maybe you've done some fixes there12:05
gordcnadya: kk. i don't think i remember it ever breaking but who knows.12:06
*** MVenesio has joined #openstack-ceilometer12:09
cdentjd__: will push it up for review soon, but I think we can get rid of TranslationHook entirely. It appears to be the result of not understanding WSGI12:11
jd__cdent: agreed on IRC, I'm not a fan either of those sync meetup, I like emails :)12:11
jd__cdent: that'd be awesome if we can remove that hook, I don't like that code12:12
*** pauldix has joined #openstack-ceilometer12:12
* gordc hopes git blame history has been erased for translationhook code12:12
gordci won't lie. i wrote it... and i don't actually know why i wrote it.12:13
cdentIt looks like it is trying to work around a workaround around a workaround etc etc12:14
jd__gordc: drunk again12:14
gordcjd__: canadians don't get drunk... sometimes.12:15
openstackgerritMerged openstack/gnocchi: rest: Add links to /v1 endpoint
openstackgerritMerged openstack/gnocchi: rest: return metrics for ../<resource_id>/metric
openstackgerritMerged openstack/gnocchi: rest: fix archive policy controller
*** yassine_ has joined #openstack-ceilometer12:31
cdentwow, this is such a mess12:35
*** afazekas_ has quit IRC12:36
*** tongli has joined #openstack-ceilometer12:38
*** jruano has joined #openstack-ceilometer12:41
*** logan2 has quit IRC13:11
*** logan2 has joined #openstack-ceilometer13:12
*** pauldix has quit IRC13:15
*** jdandrea has joined #openstack-ceilometer13:21
gordcwho has a service where i can buy stuff to break?13:25
cdentgordc: say what?13:33
*** r-mibu has quit IRC13:34
cdentjust go into the kitchen and get a stack of plates and smash them on the counter13:34
gordccdent: just accidentally ran rm against my work13:34
*** r-mibu has joined #openstack-ceilometer13:34
cdentoh no!13:34
gordci'm at office. how much does a crappy switch cost13:34
gordcmaybe i'll just bend a ethernet cable back and forth until it breaks... that'll show 'em13:35
openstackgerritChris Dent proposed openstack/aodh: Correct thread handling in TranslationHook
cdentgordc, jd__ ^13:37
gordccdent: did you trace back why the code is there? i feel like there was a bug that that fixed...13:39
cdentperhaps, but if so, that fix would never have worked in a threaded environment13:40
cdentI'll look through ceilo's history13:40
cdentmy changes should maintain the behavior that I think it was striving for13:40
cdentyeah was just going there13:41
gordcphew, not my fault.13:41
cdentthe fundamental error is that __init__ on a hook is not per request13:42
gordcah i see... that makes sense. i don't know if you want to ask bnemec what's the deal. i don't really care.13:43
cdentso that threading local is local to the thread that starts the app13:43
cdentI'm pretty confident about my change13:43
cdentit's at least not any worse13:43
gordclol kk13:43
cdentif it seems groovy I'll fix it in ceilo too13:45
*** tongli has quit IRC13:45
gordccdent: will review after i recontruct crap i deleted.13:45
cdentgordc: take heart: its like you're forced to refactor ... :) :(13:45
*** ryanpetrello has quit IRC13:47
*** pauldix has joined #openstack-ceilometer13:47
openstackgerritMerged openstack/ceilometer: Fix for rgw still throwing errors
*** edmondsw has joined #openstack-ceilometer13:49
*** ryanpetrello has joined #openstack-ceilometer13:50
openstackgerritRyota MIBU proposed openstack/aodh: Exclude event type from targets of alarm evaluator
openstackgerritRyota MIBU proposed openstack/aodh: Storage: add 'exclude' constraint to get_alarms()
*** ryanpetrello has quit IRC13:54
*** tongli has joined #openstack-ceilometer13:57
*** rbak has joined #openstack-ceilometer13:57
cdentgordc: not really worth extending the thread: but on that aodh upgrade thing: it's the disposition of the existing configuration that was the key part of those questions. Installing an aodh package will replace ceilo-alarming, but how the config is managed is an open question13:58
gordccdent: i see. yeah, that's the docco stuff ildikov would do. "note what the differences are (if any)"14:01
pradcdent, wouldnt that be updated as well? and any changes to depending configs should be handled by the deployment tools? and yea we document it accordingly14:02
gordc is the item on road map something along time lines you were thinking of?14:02
cdentprad my understand was that rpm and deb prefer not to change existing configurations? puppet-based installations, sure14:02
*** eglynn has joined #openstack-ceilometer14:03
ildikovgordc: the docco stuff comes after we answered that question cdent highlighted14:03
pradcdent, the packaging would save existing configs as .rpmsave file and drop the replacement file in14:03
pradits upto the deployment tool to update the new config accordingly14:03
cdenteven though the existing config has location specific database config?14:03
gordccdent: jd__ has a patch to reenable that... if you used alarm_connection14:05
pradrpm update will usually go conservative and stash the config regardless of whats in there14:05
pradas .rpmsave file and puts the new file as .rpmnew14:05
pradyou can use a yum-merg-conf yum plugin to merge the config automatically or do it via puppet or supporting deployment tools14:06
cdentprad in the case of aodh, there won't be a pre-existing aodh.conf but there will be the alarming-config-i-was-using-before in /etc/ceilometer14:06
cdentI guess my expectation of how an upgrade to liberty is supposed to go is a bit different from what you guys are thinking.14:07
cdentI think I upgrade and everything is still working14:07
pradcdent, well when we include aodh in the deployment we will update the puppetry accordingly to configure the conf file14:08
cdentcan we take a moment and separate ourselves from puppet, rpm and deb and any of the distros and tools14:08
cdentlet's pretend I'm a ceilometer user14:08
cdentnot a customer of some enterprise14:08
cdentI get a tarball of libery ceilometer14:09
pradthe concept is still the same.. aodh is a new module/library whatever14:09
cdentI pip install it14:09
cdentshit, alarming just broke14:09
pradand you will need to configure that accordingly14:09
pradif you're using pip to run your alarming, a breakage shouldnt be a surprise14:09
cdentNow, I'm perfectly fine with us taking the stance of being jerks about the upgrade path.14:10
gordcso again, the idea is to do that: "just upgrade".14:10
cdentBut if that is the path we are taking then we need to be _very_ verbose about it14:10
pradyea that would be a doc thingy14:10
gordcwe're doing the integration and upgrade stuff now, and it's pretty hard to say whether it's possible to 'just upgrade'14:10
openstackgerritMerged openstack/ceilometer: Enable entry points for new declarative meters
cdentwow, finally14:11
gordcif we finish all that stuff, and we find it's impossible 'just upgrade', that's where the doc comes into play14:11
openstackgerritMerged openstack/ceilometer: Convert identity, sahara and volume to meters yaml
* cdent now wonders if maybe he should have extended the email thread14:13
pradif someone is not using a distro they are use to not seeing a 'just upgrade' path .. they would likely follow a doc and go through it14:13
cdentthe same issues present themselves for people who are building the distros, of course14:14
* gordc looks at prad14:14
cdentI don't know, it just seems really weird to me to be migrating to a new project at the same time as doing a version boost14:15
cdentbut maybe I'm weird14:15
cdentI just don't want my ass bit later because we haven't hashed all this crap out now14:15
pradits indeed something to think through14:15
pradto me a seemless upgrade is a major concen for distros than some one running it locally and upgrading via pip14:16
pradbut for distros we have puppet and other tools14:16
gordccdent: sure. but i'm not sure you can say what will break on upgrade without trying it... and we're stillin process of trying it14:17
pradperhaps we can provide a bash script the user can run to merge their configs in the manual non distro case i mean14:17
pradput something in tools dir that can alleviate the pain a bit14:18
gordci don't understand the new project and version boost comment?14:18
cdentgordc: I'm trying to put myself in the mind of an operator. If I had the choice I would not want to move to aodh and to liberty in one step. I would rather upgrade to liberty and then later, at my leisure, switch to aodh.14:20
cdentI'm pretty sure we've killed that idea, though, and that's okay, just identifying my position I guess.14:20
gordccdent: upgrade not using packaging?14:23
cdent_any_ upgrade style14:23
cdentbecause I want to minimize the surface area of change14:23
cdentfeh, where's jason when I need him14:23
gordccdent: well from what prad said, they both can't exist. i'd assume the package is still the same, it just points to aodh code14:24
cdentI think the idea is that aodh packages would obsolete ceilo packages14:25
gordcwill let prad confirm... maybe i misunderstood... but i'm under impression, it's literally not possible to have both packages.14:26
cdentI'm not suggesting you would14:27
cdentan upgrade to liberty would upgrade you to ceilo-alarming (liberty)[1]14:27
cdentlater if you wante to switch to aodh, you'd install that, and it would obsolete/replaced ceilo-alarming14:27
cdent[1] currently not on the cards because we've decided to delete the code out of liberty14:28
gordc[1.1] deleted because there would be no package.14:29
gordcif 1.1 is wrong then that's another story14:29
gordcif we can build a package from it, then it sits around stale... but i'm under impression you can literally only build a package off either aodh or ceilo-alarm14:30
cdentI don't think we are quite making sense to one another.14:31
gordci think we have different understanding of what is possible and what isn't14:32
kfarhane_please can somoene tell me why i'm not seeing cpu_util metric of instances in compute node?14:32
cdentkfarhane_: we need to know what release, what virt driver, what pipeline.yaml to be able to give you a good answer to that. Are you seeing other meters from the same compute node?14:33
kfarhane_yes i'm seeing disk.root.size, disk.ephmeral.size,memory and vcpus metric only, i'm using devstack version stable/Juno, hypervisor is kvm14:35
kfarhane_cdent_: yes i'm seeing disk.root.size, disk.ephmeral.size,memory and vcpus metric only, i'm using devstack version stable/Juno, hypervisor is kvm14:36
cdentkfarhane_ have you modified pipeline.yaml or are you using the default one?14:37
kfarhane_i'm using the default one, for instances on controller node it's no problem i see all metrics, but for instances in a compute node there is only the metric i wrote14:38
openstackgerritRohit Jaiswal proposed openstack/ceilometer: Adds support for dynamic event pipeline
gordckfarhane_: the ones you listed don't come from compute agent14:39
gordcthey are built from notifications14:40
kfarhane_gordc: yes alright, but compute agent on compute nod don't poll other metric, and i don't know why?14:41
cdentjd__: I'd love to do what you suggested but unfortunately...nope.14:42
cdentkfarhane_: have you been able to look at the log files for the compute agent?14:47
gordckfarhane_: are you using the same ceilometer.conf file for both compute agents?14:47
jd__cdent: :((((14:48
gordccdent: no errors in logs when i asked yesterday14:49
kfarhane_in the log file all work fine, i can see: pollster cpu_util in the context of meter_source, and other posllster for other metric14:49
cdentjd__: it's crazy14:49
jd__cdent: am I totally off on that or is there something that prevents it?14:49
cdentthe way that stuff is implemented is...messy14:49
kfarhane_gords: yes it's the same ceilometer.conf, the only difference that on compute node section [database] in ceilometer.conf is not given14:50
gordcdo you see errors in collector? do you see anything on metering.sample queue when you run rabbitmqctl list_queues14:51
cdentjd i'll try another thing14:52
cdentjust to confirm14:52
cdentjd__: yeah, no can do14:55
cdentthere's no environ on pecan.response14:56
cdentand trying to assign to pecan.request.environ in leads to: Ensure that you're accessing `pecan.request` or `pecan.response` from within the context of a WSGI `__call__` and that `use_context_locals` = True.14:56
kfarhane_gordc: there is no erro in collector, but i'm not seeing any metering.sample when i run rabbitmqctl list_queues14:56
cdentso the hook is needed to get the required data into the environ, which the middleware can see14:57
jd__cdent: -_-14:57
cdentit's _crazy_ and very disappointing14:57
jd__missah sad14:57
kfarhane_gordc: sory sory, there is one metring.sample 0 , when i run the command rabbitmq14:57
cdentme too, but I guess it will at least work now14:58
gordckfarhane_: my only guess is you don't have any instances on the compute node...14:58
*** ildikov has quit IRC14:59
kfarhane_gordc_: sure i have an instance on compute node14:59
gordckfarhane_: if you look at compute log in addition to "polling for..." logs, you should see a dump of information related to instances14:59
gordcif you don't see a dump (assuming you have debug on), it's not finding your instance or you don't have an instance.14:59
gordcthe only other suggestion i can make is to check nova-api call to see if the polling actually makes it htere.15:01
*** ryanpetrello has joined #openstack-ceilometer15:01
gordcturn off your other compute/central agents15:02
kfarhane_gordc: can you tell me what is the write config of nova.conf and ceilometer.conf, have i to add some lines to poll metrics?15:03
kfarhane_gordc: i add theses ones: notification_driver = nova.openstack.common.notifier.rpc_notifier notification_driver = ceilometer.compute.nova_notifier, and rerun my services of ceilometer and nova, but nothing15:04
gordcthe nova_notifier is not relevant (it should be gone in juno)15:04
gordcthose items relate to notifications from nova, not polling15:05
gordcif your other compute agent works, it should be no different15:06
kfarhane_gorcd: the only compute agent which works is the one in controller node (my controller node can instanciate locally nova instances) , but compute_agent don't work in any compute_node15:08
gordccdent: if i understand your concern, it's not about upgrading but rather not upgrading. assuming there can be ceilo-alarm package built as well as an aodh package built, the question i have is what is difference between liberty ceilo-alarm vs kilo ceilo-alarm... and is 'stay on kilo ceilo-alarm' a valid answer? (since at some point that will be the answer)15:08
cdentgordc: That seems a reasonable summary.15:09
gordckfarhane_: i understand that. but in regards to conf file, they should be the same.15:09
cdentkfarhane_: are you certain that your compute node can see keystone, this was a problem I saw recently wherein the compute node was trying to use keystone on localhost on the compute node15:10
kfarhane_cdent: i think yes, because i'm able to create vm on compute node, and that require authentication keystone15:11
cdentIs that true, I thought the conductor worked over the messaging bus and keystone didn't get involved at that stage? I admit, however, I know _very_ little about nova15:12
gordccdent: cool cool. so i don't want to have both... i believe the only patch that made it to liberty was alarm trusts. is that patch alone enough to differentiate kilo alarms vs liberty alarms?15:13
openstackgerritMerged openstack/gnocchi: Implements list_resources limit/ordering
gordccdent: since we said at midcycle we should keep code, i'm ok with staying course... iff that code is still accessible. the delete asap was based on the idea that it's not accessible.15:14
cdentgordc: probably not. but is it "fair" to say to people "when you're upgrading to liberty if you don't want aodh, you need to say on kilo for ceilometer-alarming". It may very well be, just trying to understand, makes sure we're all clear etc15:14
cdentI don't personally know what the right thing to do is, nor do I understand all the options. I'm just trying to put myself in the shoes of joe-ceilometer-user and think of the questions I would have.15:15
gordc^ above condition ok? that was sort of reason why i asked yesterday what other ways is there to consume ceilometer... from packaging pov, it's not accessible... and that was only one i got.15:15
gordckfarhane_: can you check to see in nova-api, that you can see ceilometer's poll?15:18
gordcthe main part of conf you need to ensure you have is. [service_credentials]15:19
gordcthat is what ceilometer uses to connect to nova client to poll15:19
gordcunrelated -- does anyone understand yaml tags?15:20
*** r-daneel has joined #openstack-ceilometer15:22
gordcprad: i have the loop stuff working. i just need a way to say 'this field is a path' and 'this field is static'15:23
cdentgordc: i can sometimes do yaml, what's up?15:24
gordccdent: ^15:24
gordcrelates to prad's meter.yaml file15:24
cdentwhat do you mean by "path" and "static"?15:25
gordccdent: so all the unit values are static/hardcoded
gordcthe volumes would not.
*** ryanpetrello has quit IRC15:27
cdentas far as yaml is concerned those are both strings and that's all,yeah?15:27
cdentbut you want, in the yaml, to say something different?15:27
*** ryanpetrello has joined #openstack-ceilometer15:28
gordci want to know if it's a path... right now we take the string and we try to convert it to path... if we can't find path, we assume it's not a path.15:29
gordcbut i want to say 'this is a path, and not a string'15:29
gordcthe easy/stupid way is just to see if it begins with 'payload'... but i'm looking for something more clever15:31
openstackgerritChris Dent proposed openstack/aodh: Correct thread handling in TranslationHook
*** _cjones_ has joined #openstack-ceilometer15:33
pradgordc, cool15:33
*** _cjones_ has quit IRC15:33
*** _cjones_ has joined #openstack-ceilometer15:33
cdentsorry gordc, fell in a hole15:33
cdentI think you're better off doing the checking in code, not yaml, because if you put it in yaml then you're making trouble for maintainers of that file15:34
cdentin which case it sounds like what you're doing already is okay?15:34
pradgordc, couldnt we do some regex match on '.' separator for say and call it path or something better15:35
gordcit's sort of code. i'm reusing most of prad's code but using more complex jsonpath functions.15:36
pradso on math stuff i made some progress.. i added a new tag in yaml called expr a.. if expr is a string, i just eval it ..if its just a number i scale it15:37
gordcprad: yeah... i'm going to post a solution with that... we can see if it's too complex for a yaml and maybe we need to keep multi-meter payloads code based15:37
pradwe seem to be using eval all over in pipeline code..15:38
gordcwhat's this 'tag' stuff... seems like something i could ues.15:38
*** logan2 has quit IRC15:38
pradjust another field15:39
pradmay be tag is too fancy a word15:39
gordcoh. yeah. i had that idea too.lol15:40
*** logan2 has joined #openstack-ceilometer15:41
EmilienMgordc: do we have new agents/services in Liberty? I'm looking at the puppet-ceilometer status for liberty15:42
cdenttimely EmilienM15:43
EmilienMcdent: I am15:43
EmilienMcdent: and I like :)15:43
gordcprad: what would the aodh package be called? same as current ceilometer-alarm packages?15:45
EmilienMgordc: is it packaged in ubuntu too ?15:45
gordcprad: this relates to question i had, can you build both aodh and ceilometer-alarm?15:45
gordcEmilienM: it should be.15:46
gordczigo: ^15:46
EmilienMdon't ask zigo, he's working on debian/mos15:46
gordcwe don't have a release yet, so no package15:46
gordcEmilienM: who's ubuntu?15:46
EmilienMgordc: jamespage15:47
EmilienMgordc: on freenode15:47
EmilienMgordc: I tested ceilometer/liberty on ubuntu - it works15:47
EmilienMI'm just not sure if everything is installed & tested15:47
pradgordc, openstack-aodh15:47
pradgordc, we can build both15:48
pradbut cant install at the same time15:48
openstackgerritChris Dent proposed openstack/ceilometer: Correct thread handling in TranslationHook
gordcprad: ok. i was under impression only one could be built.15:49
pradwhat you mean by *built*?15:49
pradboth packages can exist in a repo , but only one can be installed15:50
gordconly one package can exist in repo15:50
pradas one obsoletes other15:50
pradwell thats a rel-eng decision, we can decide that and only have aodh in the repo15:51
*** belmoreira has quit IRC15:51
pradand when someone does yum update it will replace ceilo-alarm with openstack-aodh15:51
*** Ephur has joined #openstack-ceilometer15:51
pradtechnically there is nothing stopping us from having both is what i'm saying15:52
pradin the repo15:52
*** belmoreira has joined #openstack-ceilometer15:52
*** belmoreira has quit IRC15:52
gordci see.15:52
gordcwell then. i guess we need to kill code after L*15:52
cdentgordc: I don't follow your logic. How does the capability of making a package make difference. (I agree with you but I don't understand your reasoning)15:53
gordcthe reason for delete right away was because the code was inaccessible... but it seems that is not true anymore15:55
gordcwe can still debate whether or not we just tell people to stick with kilo if they don't want aodh... but given a choice, people will always say i want latest but safest.15:56
gordcsafest == same15:56
gordci wonder what was done when glance split out of nova.15:58
cdentthings drug on in limbo on both sides forever15:58
cdentI think that's the concern that presented for why it is better to delete15:58
gordcobviously network never died in nova15:58
gordcwell we have a delete point... i just wonder how much warning they gave when they did glance15:59
gordcasking rel mgrs regarding glance16:02
openstackgerritgordon chung proposed openstack/ceilometer: support multiple-meter payloads
*** _cjones_ has quit IRC16:04
*** _cjones_ has joined #openstack-ceilometer16:04
*** safchain has quit IRC16:06
*** eglynn has quit IRC16:18
gordcso... they split glance in b*... stopped packaging in g* and still have some random code lying around16:19
cdentyeah, so I believe the delete now idea is "just rip of bandaid now, as quickly as possible"16:21
cdentdon't prolong the pain16:21
gordcyeah. well i'm definietly not doing a 6 cycle deprecation16:25
gordcseems like M* is suffice. we already killed development pre-summit16:26
gordcseems like the least amount of time to appease the idlers.16:27
gordcif anyone has issues ^^ please raise... or comment on my reply to cdents ML16:28
*** ildikov has joined #openstack-ceilometer16:42
kfox1111if I adjust the ttl on a mongodb backed ceilometer, will it take  affect?17:00
kfox1111which agents do I need to restart too?17:01
EmilienMkfox1111: collector17:03
EmilienMmaybe alarming too17:03
*** annasort has joined #openstack-ceilometer17:05
openstackgerritChris Dent proposed openstack/aodh: Make ConnectionRetryTest more reliable
*** exploreshaifali has joined #openstack-ceilometer17:09
*** ddieterly has joined #openstack-ceilometer17:12
*** ankita_wagh has joined #openstack-ceilometer17:17
*** rjaiswal has joined #openstack-ceilometer17:19
openstackgerritMerged openstack/ceilometer: Cast Int64 values to float
openstackgerritChris Dent proposed openstack/ceilometer: Make ConnectionRetryTest more reliable
*** thorst has quit IRC17:28
*** yprokule has quit IRC17:35
cdentgordc or nadya, around?17:36
nadyacdent: yep17:37
cdentnadya: you were working on the functional test stuff yeah? are there plans already in place for dealing with the 'cover' target in tox.ini?17:37
cdentright now it still starts mongo but runs the coverage against just the unit tests17:37
cdentare we wanting to to run both function and unit tests or, just unit and it doesn't need mongo?17:38
nadyacdent: I believe the idea was to have an ability to run 'quick' unit tests without mongo start17:39
cdentyes, that's the py27 and py34 targets now17:39
nadyacdent: so mongo start may be removed when unit tests start17:39
cdentbut I'm wondering specifically about the coverage tests17:39
cdentdo we think that the coverage tests should be just unit, or unit and functional?17:40
*** IBerezovskiy has quit IRC17:42
nadyacdent: thinking...17:42
cdentcoverage is just 58% when running just the unit tests17:43
kfox1111how do I apply the ttl to old entries in mongodb?17:44
nadyacdent: what do we want from 'coverage'? Where is it used? And 58% is very bad...17:44
* cdent runs again with both17:44
cdentI think is just a useful too to find out where we need more tests (that's why I was running it when i realized this situations, I was in a test writing state of mind), so I don't think it used anywhere other than in our own work. But I'm not sure of that.17:45
nadyacdent: if we care about statistics anywhere and have 90+ now and then switch to only unit and will show 58%...17:45
cdentkfox1111: I'm trying to remember, did we talk about this a few days ago and I pointed you at some docs, or are you someone else who I haven't pointed at the docs, and thus the docs might be useful?17:46
cdentkfox1111: those docs were/are:
kfox1111yeah, it says how to set it.17:48
kfox1111but it doesn't seem to have fully kciked in. the db hasn't shrank any.17:49
cdentnadya: coverage is 84% with both functional and unit tests17:49
nadyacdent: I think to keep it as it is for now. And keep in mind that we need more unit tests17:49
kfox1111hmm... it did: "name" : "meter_ttl", "expireAfterSeconds" : 8035200,17:49
kfox1111in the db though, so its close.17:49
kfox1111maybe I have to tickle something in mongodb to get it to apply it?17:50
cdentkfox1111: I don't actually know. I thought mongo was supposed to be automatic, but I'm not sure how it works out when it wasn't set from the start.17:50
cdentlet me see if I can find some more info17:51
*** thorst has joined #openstack-ceilometer17:51
kfox1111ok. thanks. :)17:51
*** prad has quit IRC17:52
cdentkfox1111: what are you doing to determine if you expired anything or not? as far as I understand things, mongodb will not release freed disk space back to the os unless you do some extra work17:55
kfox1111just looking at the stats on the collection.17:55
*** nadya has quit IRC17:59
cdentkfox1111 you mean to choose 8035200 for the expire time? that seems a bit of a weird choice18:00
kfox1111my guess was that's about 3 months?18:00
cdentyeah, very close18:01
cdentso I guess that is what you mean18:01
kfox1111not very computer sciency looking a number. :)18:01
kfox1111we've never expired, so cutting it off there seemed ok.18:02
cdentis your installation live and you should be getting incoming samples?18:04
cdentso count in stats is going up?18:04
kfox1111trying to verify the older stuff is gone now.18:04
kfox1111its kind of slow...18:04
kfox1111good call. let me check.18:04
cdentbecause you may be losing one as a new one comes in18:04
* cdent shrugs18:04
cdentthis stuff is too hard to debug :(18:04
kfox1111ah... it is slightly less now.18:07
kfox1111so its just slow to drain out.18:07
kfox1111thanks. :)18:07
cdentI think all I did was distract you long enough for it to do something :)18:07
kfox1111could be. :)18:07
kfox1111thanks for the help.18:08
cdentyou're welcome18:08
cdenthave you heard about or had a chance to poke at gnocchi yet?18:08
cdentIt deals with the space problems in this kind of data much better18:08
cdentbut is a completely diferent data model so not a direct migration18:08
kfox1111yeah. this cloud's juno. we're going to move to kilo in a few weeks on this cloud, and then we were going to give gnocchi a try.18:09
*** jdandrea has left #openstack-ceilometer18:10
kfox1111right now I'm not sure the issue at hand would be solved by the ttl or gnocchi...18:10
kfox1111I think rabbits more the bottleneck.18:10
cdentCool. It's been under development for quite a while now but hasn't had a ton of real world testing over any significan tamount of time...18:10
kfox1111we've been requested by some of the scientists to crank up ceilometer to sample every 5 seconds.18:11
cdentwell I was just thinking the space problem and the slowness of data retrieval18:11
kfox1111Its not too happy at the moment. :)18:11
cdentgnocchi doesn't duplicate resource data with every measure, and pre-aggregates the measures18:11
kfox1111I'm guessing mongodb prioritizes writes to deletes, which is why the ttl stuf is slow to drain.18:11
kfox1111nice. that would really help space wise. :)18:11
cdenta fair guess18:11
cdenti did some pretty lame but interesting testing over the weekend:
cdent(I mention that only because I had ceilometer set to poll every 5 seconds)18:12
kfox1111cool. I'll have a read. thanks. :)18:12
cdentAnd with that I think I'm done for the evening.18:15
cdentGood night everyone18:15
*** cdent has quit IRC18:15
kfox1111good night.18:16
*** drjones has joined #openstack-ceilometer18:18
*** _cjones_ has quit IRC18:22
*** divya__ has quit IRC18:30
openstackgerritChris Dent proposed openstack/ceilometer: Don't start up mongodb for unit test coverage
*** htruta_ has joined #openstack-ceilometer18:39
*** yassine_ has quit IRC18:41
openstackgerritChris Dent proposed openstack/ceilometer: Correctly intialized olso config fixture for TestClientHTTPBasicAuth
openstackgerritgordon chung proposed openstack/ceilometer: support multiple-meter payloads
*** nadya has joined #openstack-ceilometer18:59
*** ankita_w_ has joined #openstack-ceilometer19:00
*** ankita_wagh has quit IRC19:03
*** nadya has quit IRC19:04
*** tongli has quit IRC19:28
*** exploreshaifali has quit IRC19:34
*** alexpilotti has quit IRC19:35
*** ildikov has quit IRC19:37
*** ankita_wagh has joined #openstack-ceilometer19:39
*** ddieterl_ has joined #openstack-ceilometer19:39
*** ddieterly has quit IRC19:39
*** boris-42 has joined #openstack-ceilometer19:40
*** ankita_w_ has quit IRC19:41
*** nadya has joined #openstack-ceilometer19:43
*** MVenesio has quit IRC19:45
*** rbak has quit IRC19:45
*** rbak has joined #openstack-ceilometer19:53
*** ildikov has joined #openstack-ceilometer19:57
*** gordc has quit IRC20:09
*** jruano has quit IRC20:45
*** logan2 has quit IRC20:56
*** ddieterl_ has quit IRC20:58
*** htruta_ has quit IRC20:59
*** ankita_w_ has joined #openstack-ceilometer21:00
*** htruta has quit IRC21:02
*** ankita_wagh has quit IRC21:03
openstackgerritPradeep Kilambi proposed openstack/ceilometer: Convert instance, bandwidth and SwiftMiddleware meters
*** nadya has quit IRC21:21
openstackgerritPradeep Kilambi proposed openstack/ceilometer: support multiple-meter payloads
openstackgerritPradeep Kilambi proposed openstack/ceilometer: Convert instance, bandwidth and SwiftMiddleware meters
*** htruta_ has joined #openstack-ceilometer21:37
*** logan2 has joined #openstack-ceilometer21:39
*** annasort has quit IRC21:47
*** thorst has quit IRC22:08
openstackgerritPradeep Kilambi proposed openstack/ceilometer: Add support to compute cpu meters in yaml
*** ddieterly has joined #openstack-ceilometer22:16
*** edmondsw has quit IRC22:21
*** r-daneel has quit IRC22:31
*** changbl has quit IRC22:49
*** pauldix has quit IRC22:54
*** changbl has joined #openstack-ceilometer22:56
kfox1111I've got a big backlog of unexpired tmeters in ceilometer.  plus a deadline coming up.23:08
kfox1111is it safe to djust purge all of the samples out of the meters collection?23:08
kfox1111db.meters.drop() ?23:11
*** logan2 has quit IRC23:15
*** jruano has joined #openstack-ceilometer23:30
kfox1111is it safe to djust purge all of the samples out of the meters collection?23:44
*** ankita_w_ has quit IRC23:47

Generated by 2.14.0 by Marius Gedminas - find it at!