Wednesday, 2015-09-30

*** openstackgerrit has quit IRC00:01
*** openstackgerrit has joined #openstack-ceilometer00:01
*** openstackstatus has quit IRC00:02
*** openstack has joined #openstack-ceilometer00:03
*** julim has joined #openstack-ceilometer00:08
*** devlaps has quit IRC00:09
*** ddieterly has joined #openstack-ceilometer00:34
*** lvdongbing has joined #openstack-ceilometer00:49
*** xiaofandh12 has joined #openstack-ceilometer00:50
*** lawrancejing has joined #openstack-ceilometer00:54
*** _cjones_ has quit IRC00:56
*** ankita_wagh has quit IRC01:01
marcin_kosobuckinot much help here :(01:02
*** r-mibu has joined #openstack-ceilometer01:02
openstackgerritliusheng proposed openstack/ceilometer: Rename list_events tests to list_samples tests  https://review.openstack.org/22920701:13
*** fawadkhaliq has joined #openstack-ceilometer01:33
*** Kennnan2 is now known as Kennan01:37
*** summer has joined #openstack-ceilometer01:56
*** zqfan has joined #openstack-ceilometer01:57
openstackgerritLianhao Lu proposed openstack/ceilometer: Configure collector to only record meter or event  https://review.openstack.org/21498202:00
*** agireud has quit IRC02:00
*** jfluhmann has joined #openstack-ceilometer02:07
*** ddieterly has quit IRC02:12
*** changbl has joined #openstack-ceilometer02:18
*** changbl has quit IRC02:23
*** changbl has joined #openstack-ceilometer02:25
*** ankita_wagh has joined #openstack-ceilometer02:34
*** changbl has quit IRC02:42
*** lawrancejing has quit IRC02:47
*** lawrancejing has joined #openstack-ceilometer02:48
*** fabian1 has joined #openstack-ceilometer03:09
*** fabian1 has left #openstack-ceilometer03:10
*** ddieterly has joined #openstack-ceilometer03:13
*** ddieterly has quit IRC03:18
openstackgerritJuPing proposed openstack/ceilometer: Fix the bug of "Error spelling of a word"  https://review.openstack.org/22922903:20
*** vkmc has quit IRC03:20
*** vkmc has joined #openstack-ceilometer03:24
*** agireud has joined #openstack-ceilometer03:29
*** changbl has joined #openstack-ceilometer03:31
*** ldb_ has joined #openstack-ceilometer03:38
*** changbl has quit IRC03:39
openstackgerritliusheng proposed openstack/ceilometer: Fix a mistake in a test  https://review.openstack.org/22923103:40
*** lvdongbing has quit IRC03:40
*** deepthi has joined #openstack-ceilometer03:41
*** links has joined #openstack-ceilometer03:53
*** ankita_wagh has quit IRC03:55
*** ankita_wagh has joined #openstack-ceilometer03:56
*** fawadkhaliq has quit IRC04:12
*** ddieterly has joined #openstack-ceilometer04:14
*** jfluhmann has quit IRC04:16
*** ddieterly has quit IRC04:19
*** ldb_ has quit IRC04:27
*** fawadkhaliq has joined #openstack-ceilometer04:34
*** fawadkhaliq has quit IRC04:35
*** fawadkhaliq has joined #openstack-ceilometer04:35
*** links has quit IRC04:36
openstackgerritliusheng proposed openstack/ceilometer: SQL: Fix event-list with multiple trait query filters  https://review.openstack.org/22879304:39
*** fawadkhaliq has quit IRC04:41
*** links has joined #openstack-ceilometer05:12
*** ddieterly has joined #openstack-ceilometer05:15
*** tasdomas has joined #openstack-ceilometer05:18
*** ddieterly has quit IRC05:20
openstackgerritMerged openstack/ceilometer: fix elasticsearch script reference  https://review.openstack.org/22903805:20
openstackgerritMerged openstack/ceilometer: Fix the deprecation note in meter.yaml  https://review.openstack.org/22890105:21
*** yprokule has joined #openstack-ceilometer05:24
*** alexpilotti has quit IRC05:33
*** lsmola has joined #openstack-ceilometer05:42
*** lvdongbing has joined #openstack-ceilometer05:48
openstackgerritMerged openstack/ceilometer: doc: update devstack usage  https://review.openstack.org/22123005:56
*** ddieterly has joined #openstack-ceilometer06:16
openstackgerritOpenStack Proposal Bot proposed openstack/ceilometer: Imported Translations from Zanata  https://review.openstack.org/22925406:17
*** sileht has quit IRC06:17
*** ddieterly has quit IRC06:21
*** _gryf has quit IRC06:23
*** nadya has joined #openstack-ceilometer06:28
*** sileht has joined #openstack-ceilometer06:28
*** _gryf has joined #openstack-ceilometer06:45
*** afazekas_ has joined #openstack-ceilometer07:01
*** ankita_w_ has joined #openstack-ceilometer07:03
*** ankita_wagh has quit IRC07:06
*** gzhai3 has quit IRC07:13
*** lawrancejing has quit IRC07:13
*** khushbu_ has joined #openstack-ceilometer07:14
*** gzhai has joined #openstack-ceilometer07:14
*** lawrancejing has joined #openstack-ceilometer07:15
*** ddieterly has joined #openstack-ceilometer07:16
*** gzhai1 has joined #openstack-ceilometer07:21
*** ddieterly has quit IRC07:22
*** Kennan has quit IRC07:23
*** summer has quit IRC07:23
*** gzhai has quit IRC07:24
*** xiaofandh12 has quit IRC07:24
*** alexpilotti has joined #openstack-ceilometer07:24
*** summer has joined #openstack-ceilometer07:25
*** xiaofandh12 has joined #openstack-ceilometer07:25
*** Kennan has joined #openstack-ceilometer07:25
*** shardy has joined #openstack-ceilometer07:29
*** khushbu_ has quit IRC07:35
*** safchain has joined #openstack-ceilometer07:40
*** afazekas_ has quit IRC07:40
*** alexpilotti has quit IRC07:58
*** ankita_w_ has quit IRC08:05
*** ankita_wagh has joined #openstack-ceilometer08:05
openstackgerritMerged openstack/ceilometer: Fix the bug of "Error spelling of a word"  https://review.openstack.org/22922908:08
*** ildikov has quit IRC08:10
openstackgerritMerged openstack/ceilometer: Fix a mistake in a test  https://review.openstack.org/22923108:12
*** alexpilotti has joined #openstack-ceilometer08:13
*** Ala has joined #openstack-ceilometer08:14
*** yassine__ has joined #openstack-ceilometer08:15
*** ankita_wagh has quit IRC08:15
*** khushbu_ has joined #openstack-ceilometer08:15
*** ankita_wagh has joined #openstack-ceilometer08:15
*** ddieterly has joined #openstack-ceilometer08:17
*** ankita_wagh has quit IRC08:20
*** ddieterly has quit IRC08:22
*** shardy_ has joined #openstack-ceilometer08:49
*** shardy has quit IRC08:51
*** agireud has quit IRC08:54
*** shardy_ has quit IRC08:55
*** shardy has joined #openstack-ceilometer08:55
*** cdent has joined #openstack-ceilometer08:57
*** deepthi has quit IRC09:00
*** nijaba has quit IRC09:01
*** khushbu_ has quit IRC09:01
*** khushbu has joined #openstack-ceilometer09:02
*** alexpilotti has quit IRC09:04
openstackgerritJulien Danjou proposed openstack/gnocchi: Create signature of measures_report in base class  https://review.openstack.org/22912409:06
*** nijaba has joined #openstack-ceilometer09:08
*** deepthi has joined #openstack-ceilometer09:09
cdentthanks for that jd__09:09
jd__np :)09:10
cdentbut can you explain why if the implementation in the subclass is not static and presumably never would be the super is?09:10
jd__the static/class methods are tight to the current class09:10
jd__there no link to the general class inheritance hierarchy in Python09:11
cdentYeah, I get that, I just mean in terms of the base being representative of the interface (as a human aid)09:11
jd__the human aid is the parameters and the return value (documented in the doc string), static or not static is not representative09:12
jd__you could have a static implementatino of the method in the driver09:12
jd__even if the base class is not declared static like you did09:12
cdentcool, thanks09:12
jd__I know it's really weird, but it's how Python works09:12
jd__the decorators are just used to indicate if the method should be bound to an instance or not internally09:13
jd__so it's really class-local09:13
cdentI think once the influxdb stuff is back to almost working, the experimental influxdb is going to break because of measures_report needing to be implemented09:13
cdentbut that's some distance away as the other tests are still messed up09:14
jd__you can implement it as "pass" I guess09:14
jd__yeah no progress on the other tests?09:14
cdentI think the issue is in get_measures09:15
cdentbut I haven't decoded it enough yet to be certain09:15
cdenthowever apart from that if I run the tests enough times I can tickle the timeout bug09:16
cdentso "experimental" seems wise09:16
*** ddieterly has joined #openstack-ceilometer09:18
*** ddieterly has quit IRC09:23
*** r-mibu has left #openstack-ceilometer09:23
*** xiaofandh12 has quit IRC09:33
*** lvdongbing has quit IRC09:41
*** khushbu has quit IRC09:44
*** khushbu has joined #openstack-ceilometer09:47
*** liusheng has quit IRC09:47
*** lawrancejing has quit IRC09:54
*** summer has quit IRC09:55
*** khushbu has quit IRC10:03
*** khushbu has joined #openstack-ceilometer10:11
*** prnk28 has joined #openstack-ceilometer10:14
*** khushbu has quit IRC10:18
*** ddieterly has joined #openstack-ceilometer10:19
*** ildikov has joined #openstack-ceilometer10:20
*** ddieterly has quit IRC10:23
*** ildikov has quit IRC10:33
*** deepthi has quit IRC10:35
*** shardy_ has joined #openstack-ceilometer10:39
*** shardy has quit IRC10:41
*** shardy_ has quit IRC10:45
*** shardy has joined #openstack-ceilometer10:45
*** ildikov has joined #openstack-ceilometer10:49
ityaptinjd__, cdent: Hi!10:59
cdenthi ityaptin, how goes?11:00
ityaptinI also tinker with influxdb timeout and timeouts this week.11:01
cdentany progress so far?11:02
jd__cool11:02
cdentbrb11:02
*** lawrancejing has joined #openstack-ceilometer11:02
ityaptinI making two things - 1. implement continuous queries for aggregation measures according to AP and getting aggregated data via get_measures 2. writing raw_data to "raw_data" measurement with tags like (metric_id, metric_name, resource_id, resource_type) if it exists11:05
ityaptinI don't sure what it fixes issue with writing to influxdb but in tests I didn't see timeouts yesterday11:06
ityaptincdent: Today-tommorow I want to send POC to review (I need to check correctness of output data). could you also try it on your env?11:08
ityaptinjd__, cdent: does wal option for 0.9.3 also leave much to be desired for fast writing?11:09
ityaptinjd__: I see what you patched tox configuration of influxdb with this option.11:10
jd__ityaptin: did I? I copied something I guess from cdent or from the default config11:10
ityaptinjd__ : "wal-dir" in influxdb config generation. According to the description wal should compile different measures to one batch and flush it every couple of seconds11:12
jd__ityaptin: yeah that option is now mandatory11:13
jd__InfluxDB does not start without it11:13
jd__I just set a default, that's it11:13
ityaptinjd__: Maybe we can play with wal options for avoid the writing issue? now it have many different option for batching data: https://github.com/influxdb/influxdb/blob/c5a82e062b42f4356a9d5229b95d13f94d9fe40b/etc/config.sample.toml#L4511:15
ityaptincorrect - https://github.com/influxdb/influxdb/blob/master/etc/config.sample.toml#L4511:16
jd__ityaptin: I have no clue, I would expect it to work by default11:16
*** ddieterly has joined #openstack-ceilometer11:20
ityaptinHow I see in relative bug  cdent also trying to improve size of batched measures and it doesn't fix this issue completely(11:20
*** lvdongbing has joined #openstack-ceilometer11:23
*** ddieterly has quit IRC11:24
ityaptinjd__: And another question should I write a spec for InfluxDB CQ aggregation11:27
cdentsorry, ityaptin, jd__: got pulled away for a bit there. I agree with jd that it seems like it ought to "just work" so we are somehow doing something wrong or influxdb is just busted11:27
cdentI can't imagine it is the latter because other people do use it successfully and we see the problems in _very_ light loads11:27
*** khushbu_ has joined #openstack-ceilometer11:30
*** lvdongbing has quit IRC11:33
*** nadya has quit IRC11:34
*** julim has quit IRC11:34
*** prnk28 has quit IRC11:35
*** julim has joined #openstack-ceilometer11:39
*** gordc has joined #openstack-ceilometer11:42
ityaptincdent: generally I agree with you but think that we will need to tune InfluxDB configuration in future. About our wrong working with InfluxDB  - maybe we should to trying batch data before writing it to influx?  How other guys made.11:44
*** julim has quit IRC11:44
openstackgerritMerged openstack/ceilometer: Imported Translations from Zanata  https://review.openstack.org/22925411:45
openstackgerritMerged openstack/gnocchi: Create signature of measures_report in base class  https://review.openstack.org/22912411:45
openstackgerritMerged openstack/ceilometer: Rename list_events tests to list_samples tests  https://review.openstack.org/22920711:45
ityaptinFor batching all measures from one polling packet. now we trying to write 1 measure by time according to regular ceilometer polling behavior.11:45
ityaptinbecause we have 1 measure for metric due the 1 polling interval generally11:46
cdentityaptin: I don't think that will have any impact on the timeout problems: I'm able to cause the timeout problems when running the tests11:47
cdentbut only intermittently11:47
ityaptincdent: what is the duration of timeout? 1 minute?11:48
cdentthe default that is being tickled is 5s. I've raised it to 15s in the past and it still happens.11:48
cdentI don't think it is really a timeout so much as a deadlock of some kind11:48
ityaptinthis timeouts happen in tox -e <influxdb-job> ?11:50
cdentonly sometimes11:50
cdentI can get it to happen reliably manually (running a devstack with ceilometer+gnocchi) and irregularly when tox -e $influxdb11:50
ityaptincdent: what version of influxdb you run?11:51
*** jruano has joined #openstack-ceilometer11:52
cdentthis has happening with 0.9.2, 0.9.3, 0.9.4 and yesterday with the latest nightly11:52
ityaptincdent: ok) I'm trying to get timeout with `tox -e $influxdb` for 0.9.4 and trying research the reason today.11:53
*** thorst has joined #openstack-ceilometer11:53
cdentWhat I was doing yesterday was just running `tox  -epy27-postgresql-influxdb -- TestStorageDriver` over and over11:54
cdentand every now and again it would timeout11:54
cdentunfortunately there are several tests in there that also fail (return the wrong data) so it is hard to be clear on what's going on11:54
cdentI tried to figure out why the wrong data was coming back but didn't figure it out. I think, but am not sure, that it is a problem in get_measures11:55
ityaptincdent: also problem may be in influxdb python package. Yesterday i find that writing points via `time: datetime.datime()` or `time: "2015-09-29T00:00:00.123143+00:00"`does not work for 'n' time precision(12:00
*** Ephur has joined #openstack-ceilometer12:00
ityaptincdent: In previous version it works12:00
ityaptinconvert_timestamp doesn't work good, how i understand12:01
ityaptinbut it doesn't raise any errors12:01
cdent:(12:01
ityaptinso, I replace it to int timestamp in my code, maybe it also affect12:02
*** julim has joined #openstack-ceilometer12:03
ityaptinlet me try it today and I will fix known (for me) data writing/getting issues. Also, I want add Continuous Queries very strong, because it allow to avoid  hacks for aggregation data in realtime. Influxdb queries is not designed for our realtime queries12:05
*** htruta` is now known as htruta12:05
cdentgood luck!12:06
*** ildikov has quit IRC12:06
*** MVenesio has joined #openstack-ceilometer12:10
ityaptincdent: I will prepare a etherpad with my thoughts and design of CQ for Gnocchi12:14
ityaptincdent: or we have a Gnocchi spec space?12:14
cdentityaptin: that makes sense. For our own sanity it would be great if we could solve (or at least get a clear picture of) the problem with timeouts before adding any of the CQ stuff12:14
openstackgerritJulien Danjou proposed openstack/gnocchi: storage: remove index from measures_report()  https://review.openstack.org/22939112:15
openstackgerritJulien Danjou proposed openstack/gnocchi: rest: export overall status  https://review.openstack.org/22939212:15
*** prnk28 has joined #openstack-ceilometer12:17
*** ddieterly has joined #openstack-ceilometer12:20
*** ddieterly has quit IRC12:26
*** ildikov has joined #openstack-ceilometer12:30
*** julim has quit IRC12:30
*** jruano has quit IRC12:31
openstackgerritJulien Danjou proposed openstack/python-gnocchiclient: Remove upper cap on Sphinx  https://review.openstack.org/22939712:33
*** julim has joined #openstack-ceilometer12:34
*** Kennan has quit IRC12:38
*** Kennan has joined #openstack-ceilometer12:39
*** ddieterly has joined #openstack-ceilometer12:40
*** ala_ has joined #openstack-ceilometer12:49
*** Ala has quit IRC12:51
*** edmondsw has joined #openstack-ceilometer12:54
openstackgerritJulien Danjou proposed openstack/gnocchi: rest: export overall status  https://review.openstack.org/22939212:54
openstackgerritMerged openstack/ceilometer: SQL: Fix event-list with multiple trait query filters  https://review.openstack.org/22879312:54
*** ddieterly has quit IRC13:00
openstackgerritJulien Danjou proposed openstack/python-gnocchiclient: Add support for /v1/status  https://review.openstack.org/22941013:01
*** nadya has joined #openstack-ceilometer13:04
*** IBerezovskiy has joined #openstack-ceilometer13:05
*** julim has quit IRC13:13
*** julim has joined #openstack-ceilometer13:15
*** julim has quit IRC13:16
*** ildikov has quit IRC13:24
*** julim has joined #openstack-ceilometer13:27
gordcjd__: question raised last night, are any plans to have default archive policy set by gnocchi? right now it seems to be set by ceilometer.13:27
jd__gordc: you just have to set an archive policy rules for *13:28
gordchttps://github.com/openstack/ceilometer/blob/master/ceilometer/dispatcher/gnocchi.py#L50-L5313:28
gordcjd__: what's '*'?13:28
jd__though I'm not sure it'd work as expected as I'm not sure there is any guarantee on the order of evaluation13:29
jd__http://docs.openstack.org/developer/gnocchi/rest.html#archive-policy-rule set metric_pattern to *13:29
*** alejandrito has joined #openstack-ceilometer13:30
gordcoh. yeah, i guess the issue was it'd be nice to have a path where we didn't have to define archive policy explicitly.13:31
jd__you always have to define an archive policy for a metric, either explicitely, either by a rule13:31
jd__I just created https://bugs.launchpad.net/gnocchi/+bug/1501344 for that ordering issue13:31
openstackLaunchpad bug 1501344 in Gnocchi "Evaluation of archive_policy_rules is unspecified" [Medium,Triaged]13:31
gordci think the comment was right now we have define policy, and set it to default in ceilometer. it'd be easier if there was a default policy provided by gnocchi if not explicit stated.13:32
*** ddieterly has joined #openstack-ceilometer13:32
*** ddieterly has quit IRC13:33
*** ddieterly has joined #openstack-ceilometer13:33
jd__gordc: there's no archive policy created by default, so we can't really set any default13:34
jd__default=??? :)13:34
gordcjd__: is archive_policy_rule overridden when you specify a policy explicitly? or is it AND?13:34
jd__gordc: it's overriden13:34
jd__the rule is just used when you set nothing in the creation request13:34
*** khushbu_ has quit IRC13:35
gordcright. can we have a default rule?13:35
gordcthe usability issue right now is that we need to sync default in two places. ie defined the default policy in ceilometer and then ensure that default exists in gnocchi.13:37
gordcs/default exists/policy exists/13:38
jd__what you need is to 1. create an archive policy in Gnocchi 2. set to use that as default in Ceilometer13:40
jd__I'm not sure how to cut that13:40
gordchave a default rule in gnocchi which applies to all metrics... not possible?13:42
alejandritomorning, i have an ... existencial GNOCCHI question. i need to know how to do things "THE GNOCCHI WAY". lets suppose i have the "cpu.util" metric and for billing , showback and realtime metrics matters y need to store, i dunno, like a minnimum of 6 granularities for a metric, since i know i cant asociate more than one archive policy per metric, shoul i create a single policy with ALL my needed granularities ?  and afte13:45
alejandritor that ... if six months later y say "oh, i forgot to add one granularity because my bussiness changed" do i really need to delete all the data that metric got till that time, just because i need to add a new policy with all the previous granularities plus the new one ? again, i want to read from you whats the GNOCCHI WAY of doing things13:45
*** julim has quit IRC13:45
cdentTheGnocchiWay™13:45
* cdent looks at jd__ 13:45
*** nihilifer has quit IRC13:47
*** nihilifer has joined #openstack-ceilometer13:47
*** nihilifer has quit IRC13:47
*** nihilifer has joined #openstack-ceilometer13:47
jd__gordc: so you want: 1. create an archive policy in Gnocchi 2. Set the default archive policy rule in Gnocchi? We can do that, but if someone forget to do 2, the Ceilometer collector is going to blow – though I guess having a wrong default in current 2. also explodes13:48
*** dan-t has joined #openstack-ceilometer13:48
jd__alejandrito: so for now, the answer is yes to have to do that way13:49
jd__alejandrito: technically, we could make this use case simpler I guess, but we don't have such a feature yet13:49
jd__alejandrito: the archive policies are never meant to be modified, but we could offer an API call that would COPY a metric to a new metric with a different archive policy, considering the archive policies has some overlap in their definition13:50
*** julim has joined #openstack-ceilometer13:51
*** sergio_nubeliu has joined #openstack-ceilometer13:51
alejandritojd__, hooooooooooooooooooo !!!! that would be the ultimate solution ! would be amazing13:52
alejandritojd__, since gnocchi will be heavily used for billuing purposes also, and "granularities" will vary constantly on the same metric drive by "bussiness"13:53
alejandritogordc, thanks so much to engage with jd__ the default policy that we talked yesterday ^_^13:53
jd__alejandrito: yeah – keep in mind it's gonna likely be possible to downscale a metric (e.g. going from 5min to 1hour) but not to upscale13:53
jd__I also want to explore the road of upscaling by offering a no-aggregation storage option but that's still a bit more far on my roadmap13:54
*** pradk has joined #openstack-ceilometer13:54
alejandritojd__, we feel that if downscale its possible for existing data13:57
alejandritojd__, but upscale its for actual / future data, you have a lot fixed there already13:57
jd__alejandrito: that should be possible indeed13:58
jd__you'd need some tool to batch-apply that then, but nothing imposible13:58
*** julim has quit IRC14:00
alejandritojd__, for sure ^_^ what im wondering is how can we do to define how important is to have this done, and how can we help ? ( code included )14:02
jd__alejandrito: I'm gonna open bugs in Launchpad for a starter :) then I can't give you any deadline but we can try to split the work in pieces and move step by step14:04
*** lvdongbing has joined #openstack-ceilometer14:05
alejandritojd__, that would be amazing, want me to open the bugs ?14:05
jd__alejandrito: I'm gonna do that, just proof-read them :)14:06
*** lvdongbing has quit IRC14:06
alejandritojd__, terrific14:07
jd__alejandrito: https://bugs.launchpad.net/gnocchi/+bug/1501352 that's a first one14:07
openstackLaunchpad bug 1501352 in Gnocchi "Allow to copy a metric with a different archive policy" [Wishlist,Triaged]14:07
jd__you can subscribe to it14:07
*** julim has joined #openstack-ceilometer14:07
jd__for upscaling I'm not sure how to do that currently, because having a definition that has no measures in an archive is going to give bad results in e.g. aggregation14:08
alejandritojd__, starting with the copying policy we feel its an 100% improvement ^_^14:09
alejandritojd__, true that up/down scaling can take more time14:09
jd__alejandrito: https://bugs.launchpad.net/gnocchi/+bug/150135514:11
openstackLaunchpad bug 1501355 in Gnocchi "Allow to store measures with no aggregation" [Wishlist,Triaged]14:11
alejandritojd__, the bug/feature its perfect14:11
alejandritojd__, will read the secconf one14:11
jd__feel free to add comments if you need14:12
*** j_king has joined #openstack-ceilometer14:12
*** shardy_ has joined #openstack-ceilometer14:13
gordcjd__: maybe we can hide the default in client. if you post a metric without an archive policy, it will use default (and if that policy don't exist, it will create it also)... i think the key is that we shouldn't *need* to define a default in ceilometer14:14
*** shardy has quit IRC14:14
jd__gordc: so we can make the default to None in Ceilometer and just send nothing by default, hoping there's an archive policy rule in Gnocchi that will match, how does that sound?14:16
*** j_king has left #openstack-ceilometer14:17
gordcjd__: and if there's no rule/policy it explodes?14:17
*** shardy_ has quit IRC14:18
*** shardy has joined #openstack-ceilometer14:19
alejandritojd__, will add a comment to the seccond bug, that its the most time consuming14:19
jd__gordc: if there's no rule it explodes14:20
gordcjd__: a little better i guess. was hoping to have a rule (and corresponding policy) just baked in...14:26
gordcbut i guess ultimately there's no ideal rule/policy so that should probably configured regardless.14:26
*** jruano has joined #openstack-ceilometer14:27
openstackgerritPradeep Kilambi proposed openstack/ceilometer: Remove deprecated existence meters  https://review.openstack.org/22856314:29
openstackgerritPradeep Kilambi proposed openstack/ceilometer: Remove deprecated network existence meter  https://review.openstack.org/22857914:29
jd__gordc: that's the root of the problem. we hardly can create a default archive policy and let people store metrics and then realise the 1 hour average is not fine enough…14:29
alejandritojd__, gordc we will love to have the latest news about the "default archive policies" if we miss something and want to share it with us / want us to test something right away, you can reach us all at infra@nubeliu.com if anything is needed ^_^14:30
*** prnk28 has quit IRC14:30
gordcjd__: ack. yeah. for devs it's a lot easier but reality probably not super useful.14:31
gordcalejandrito: just talking it over right now... ^^14:32
gordcalejandrito: the problem with default is implicit defaults may not be very useful in reality.14:33
gordceveryone has different requirements.14:33
gordcultimately it might make sense for users to create a policy and policy rule and use that as their default.14:34
alejandritogordc, absolutely agreed, its a user's dutty to define not only the policies and the rules, but the default/fallback policy ^_^14:36
alejandritogordc, jd__ and as a default policy, it should be on gnocchi's policy/rule side14:38
alejandritogordc, not dispatcher14:38
alejandritogordc, IMHO14:39
jd__we can set the default value to None in dispatcher14:39
gordcalejandrito: makes sense to me.14:39
*** agireud has joined #openstack-ceilometer14:41
openstackgerritJulien Danjou proposed openstack/ceilometer: gnocchi: set the default archive policy to None  https://review.openstack.org/22946114:42
jd__gordc: alejandrito ^^14:42
*** shardy_ has joined #openstack-ceilometer14:42
*** rakhi has joined #openstack-ceilometer14:43
*** shardy has quit IRC14:44
gordcjd__: here's your bug https://bugs.launchpad.net/ceilometer/+bug/150137214:44
openstackLaunchpad bug 1501372 in Ceilometer "archive_policy in gnocch dispatcher should be optional" [Medium,Triaged] - Assigned to Julien Danjou (jdanjou)14:44
gordcmaybe we can include this in liberty-rc2... seems safe enough14:45
alejandritojd__, gordc that's amazing BUT, how can we define a default policy from gnocchis side ? its not currently supported right ? i mean ... a policy that matches if none matches and fallbacks to an archive policy14:45
openstackgerritJulien Danjou proposed openstack/ceilometer: gnocchi: set the default archive policy to None  https://review.openstack.org/22946114:45
gordcalejandrito: you can define an archive_policy_rule14:45
jd__alejandrito: create an AP rule that matches the metric_pattern '*'14:45
gordchttp://docs.openstack.org/developer/gnocchi/rest.html#archive-policy-rule14:45
*** shardy_ has quit IRC14:48
alejandritojd__, i was hoping for you to tell me that ... ( and we tested this yesterday )if you define a set of rules, but you define a rule with *, all other rules becomes null .. and every new metric ALLWAYS goes to * despite matching other rules14:48
*** shardy has joined #openstack-ceilometer14:49
alejandritojd__, gordc like there's no order on analyzing rules14:49
jd__alejandrito: that's bug https://bugs.launchpad.net/gnocchi/+bug/150134414:50
openstackLaunchpad bug 1501344 in Gnocchi "Evaluation of archive_policy_rules is unspecified" [Medium,Triaged]14:50
jd__:-)14:50
jd__easy to fix if you want to try14:50
*** rakhi has quit IRC14:51
*** rakhi has joined #openstack-ceilometer14:54
gordcjd__: what's your expected result for multiple matches? apply all matching rules rather than one?14:56
alejandritojd__, what i meant, and i dunno if we are 100% synced, its that other that if .* gets analyzed or not, what i meant, its that if you have a set of rules like cpu.*, net.*, disk.* and then i define a rule that just says '*' all other rules are nullified and not analyzed just because a rule that says JUST * exists14:56
gordcalejandrito: i'm leaning towards applying all matching rules as the fix ^... what's your expected behaviuor?14:57
jd__gordc: nah it should be first match applied, but you do not know what first is right now14:57
jd__alejandrito: yeah that's what the bug is about14:57
jd__alejandrito: you cannot know if '*' is going to be evaluated before disk.*14:58
jd__alejandrito: same problem is going to appear for disk.* vs disk.io.*14:58
gordcjd__: when would you ever know what the first match is?14:58
jd__gordc: if you know in which order they are evaluated14:58
jd__e.g. alphabetical order of the rules (the rules have names)14:58
gordcjd__: that's a very tricky way to figure out ordering...14:59
alejandritogordc, jd__ we are desussing it here also ^_^ an alphabetical order of the rules names could be a solution15:00
alejandritogordc, jd__ OR, maybe adding a policy_rule_number field to every rule15:01
*** r-daneel has joined #openstack-ceilometer15:01
gordcif you create a new rule, you'd need to rename rule(s) to get it in proper order? i don't ilke that.15:01
jd__gordc: feel free to propose something better15:02
jd__there's a bug opened :)15:02
gordcalejandrito: re: policy_rule_number, what if you have policies that are not related.15:02
gordcjd__: discussin it here now :)15:02
alejandritogordc, jd__ discussing it internally ^_^ hahaha15:03
alejandritogordc, jd__ we are thinking that if rule pattern15:05
alejandritogordc, jd__ just supports * wildcard, and aphabetical order (DESC) by the pattern and not by the name could work15:06
gordcalejandrito: what's a scenario where you match against multiple rules?15:08
gordcerr nevermind. just read my quesion.lol15:09
gordci asnwered it myself.15:09
alejandritogordc, hahah15:09
gordcyeah i guess if the pattern works, that makes sense.15:10
gordcpattern ordering*15:10
alejandritogordc, jd__ we just tested it, and works like a charm ^_^15:11
gordc:)15:11
* gordc waits for the review.15:11
alejandritogordc, could be a good and cheap approach15:11
alejandritogordc, to solve the DEFAULT rule15:12
gordcseems like it15:12
*** prnk28 has joined #openstack-ceilometer15:12
alejandritogordc,  will add the comment on the bug15:13
gordccool cool15:14
jd__makes sense15:14
jd__alejandrito: you send a patch? :)15:14
alejandritojd__, we can try ^_^ can i send you a pastebin with the result ?15:16
alejandritojd__, is that ok ?15:16
jd__alejandrito: it's better than nothing :) paste it in the Launchpad bug15:16
alejandritojd__, just because i dont quite understand the whole process to contrinute the right way :O15:17
jd__alejandrito: sure, that's a good opportunity to learn it, but don't worry we'll fix it anyway :)15:17
gordcalejandrito: alejandro [ at ] nubeliu? we can add you as co-author if you're not sure how to contribute yourself15:20
*** rakhi has quit IRC15:21
alejandritogordc, infra@nubeliu.com would be great ^_^ we are on it now !15:21
alejandritogordc, thx so much15:21
*** links has quit IRC15:21
alejandritogordc, sorry, dev@nubeliu.com ^_^15:22
gordcthat's a registered openstack account?15:22
*** prnk28 has quit IRC15:24
*** rakhi_ has joined #openstack-ceilometer15:26
*** julim has quit IRC15:28
*** rakhi_ has quit IRC15:29
*** julim has joined #openstack-ceilometer15:31
*** rakhi_ has joined #openstack-ceilometer15:32
*** julim has quit IRC15:36
*** julim has joined #openstack-ceilometer15:37
*** adurbin_ has joined #openstack-ceilometer15:46
*** IBerezovskiy has quit IRC15:57
alejandritohmmm , nope ... alejandro@nubeliu.com is15:57
*** ViswaV has joined #openstack-ceilometer15:58
*** ddieterl_ has joined #openstack-ceilometer15:58
*** ddieterl_ has quit IRC15:58
*** ddieterly has quit IRC15:59
*** ddieterly has joined #openstack-ceilometer15:59
*** ViswaV has quit IRC16:02
*** ViswaV has joined #openstack-ceilometer16:03
*** jwcroppe has quit IRC16:05
*** ildikov has joined #openstack-ceilometer16:09
*** ddieterly has quit IRC16:13
*** ddieterly has joined #openstack-ceilometer16:14
*** adurbin__ has joined #openstack-ceilometer16:17
*** adurbin_ has quit IRC16:17
ityaptincdent: Hi, again!16:20
cdenthello16:20
ityaptincdent: I run tests on my private configured InfluxDb and watched wonder picture: After `n ` test iterations writing one any measure to `gnocchi` database take a long-long time.16:23
gordcsileht: is the zmq driver some special case? https://review.openstack.org/#/c/227913/16:23
cdentityaptin: yeah that sounds familiar16:23
ityaptincdent: even we make it via api from postman16:23
*** ityaptin has quit IRC16:25
*** ankita_wagh has joined #openstack-ceilometer16:25
*** ityaptin has joined #openstack-ceilometer16:25
*** ankita_wagh has quit IRC16:25
*** ankita_wagh has joined #openstack-ceilometer16:26
*** ala_ has quit IRC16:26
ityaptincdent:  I have a unconfirmed suspicions that it caused by aged dates like '2014-01-01T00:00:00'. I'm going to run writing data with scatter up to 2 years and check writing rate.16:27
alejandritogordc, jd__ fix proposed for the rules matching ^_^ https://bugs.launchpad.net/gnocchi/+bug/150134416:27
openstackLaunchpad bug 1501344 in Gnocchi "Evaluation of archive_policy_rules is unspecified" [Medium,Triaged]16:27
alejandritogordc, jd__ tested and working OK when creating metrics that match, and metrics that dont match and goes to * rule ^_^16:28
gordcalejandrito: nice.16:29
gordcthanks for the contribution!16:29
alejandritogordc, hope it helps !16:29
*** rakhi_ has quit IRC16:29
cdentityaptin: I wondered about that too16:34
cdentbut will be distressed if that's the case16:34
cdentalso it doesn't explain the problem seen when just using it for real in devstack16:35
*** ViswaV has quit IRC16:35
cdentityaptin: yeah, given that it fails in regular use (with contemporary times) something else must be up16:38
*** _cjones_ has joined #openstack-ceilometer16:42
*** _cjones_ has quit IRC16:42
*** _cjones_ has joined #openstack-ceilometer16:42
*** lsmola has quit IRC16:44
*** jwcroppe has joined #openstack-ceilometer16:44
gordcjd__: i'm going to take the bug and build a test. please tell me to stop if you want it asap.16:45
jd__gordc: #1501344 ?16:45
gordcyep16:45
gordcwant it asap == you want the patch right away. not tell me if you want it... but that too.16:46
jd__gordc: no go ahead16:47
jd__that's a good idea :)16:47
gordcjd__: figured i'd start small.16:47
*** devlaps has joined #openstack-ceilometer16:50
*** cdent has quit IRC16:52
*** harlowja has quit IRC16:52
*** yprokule has quit IRC16:53
*** vkmc is now known as vkmc-afk17:03
*** yassine__ has quit IRC17:05
marcin_kosobuckiCould you guys give me some pointers to how to setup ceilometer alarm based on VM instance status (running, error) etc. Essentially I would like to have some auto-healing of heat stack.17:07
*** nadya has quit IRC17:08
*** ankita_wagh has quit IRC17:11
*** safchain has quit IRC17:11
marcin_kosobuckianybody?17:13
alejandritomarcin_kosobucki, with ceilometer alarms or with aodh ?17:15
marcin_kosobuckiceilometer17:16
marcin_kosobuckialarms17:16
marcin_kosobuckiwhat is aodh17:19
*** lawrancejing has quit IRC17:23
alejandritomarcin_kosobucki, http://docs.openstack.org/developer/ceilometer/architecture.html#alarming-service17:23
alejandritogordc, jd__ do you remember the review link so i can apply the change where the dispatcher passes None to gnocchi in the archive policy ?17:26
*** ankita_wagh has joined #openstack-ceilometer17:27
gordcalejandrito: https://review.openstack.org/#/c/229461/17:33
alejandritogordc, thx !17:34
gordcmarcin_kosobucki: alarming functionality in ceilometer repo is deprecated... it was moved to it's own service: Aodh17:34
*** shakamunyi has quit IRC17:35
*** vkmc-afk is now known as vkmc17:38
alejandritogordc, jd__ we are seeing that the timestamp on measures get doesnt respond not even with the agents time, collector time nor the gnocchi api server, where is the timestamp supposed to come from ?17:38
gordcmarcin_kosobucki: this might help http://myopensourcelife.com/2014/09/13/autoscaling-in-openstack-using-heat-and-ceilometer-part-1/17:38
*** vkmc has quit IRC17:38
*** vkmc has joined #openstack-ceilometer17:38
gordcalejandrito: depends what the metric is.17:39
alejandritogordc, wow, didnt expect that !17:40
alejandritogordc, that means ?17:40
gordcpolled data usually corresponds to agent time. notification data might be when it's received, or when the service generated notification, or some time value the service gives us.17:40
alejandritogordc, we know that not only agents and notifications, but servers composing the solution have the same timezone / time17:42
nijabagordc: 12 entry so far in the survey, some good stuff17:43
nijabaI will not be able to join the meeting tomorrow, so wanted you to have an update17:44
gordcnijaba: cool! i think we'll leave it running for a while.17:44
gordcare the results only visible to creator?17:45
nijabaAnyone with edit rights can see the results17:45
nijabaI think you have those17:45
gordcnijaba: oh ok. i'll give it a try and will let you know if i have issus.17:45
nijabagordc: you, julien and I have full access17:46
gordcnijaba: cool cool.17:46
nijababeware that any change to the result spreadsheet are for good, so avoid that ;)17:46
nijabagordc: it could be a good thing to ask people to relay the information. I have not seen many tweets about it apart from mine, for example17:47
gordcnijaba: good to know. i'll try to click as little as possible when i check.17:48
gordcnijaba: sure. i passed it on to product wg, they will mentioned via their twitter as well.17:50
nijabagood to hear.  thanks17:50
*** nadya has joined #openstack-ceilometer17:56
*** shakamunyi has joined #openstack-ceilometer18:01
*** ViswaV has joined #openstack-ceilometer18:05
*** ViswaV has quit IRC18:09
*** ViswaV has joined #openstack-ceilometer18:10
marcin_kosobuckigordc: Autoscaling based on cpu_util I can do easy, that is not a problem, what I want is to rebuild a VM if lets say its Compute hardware is in downgraded state, similar to what AWS offers18:13
*** harlowja has joined #openstack-ceilometer18:13
gordcmarcin_kosobucki: is 'downgraded state' a hypothetical or real value?18:15
*** openstackgerrit has quit IRC18:16
gordcyour question might better answered if asked in #heat if it's specifically about how to rebuild based on an alarm.18:16
*** openstackgerrit has joined #openstack-ceilometer18:16
gordcceilometer captures meters/metrics, aodh defines rules and triggers, heat acts on the alarms and orchestrates18:18
marcin_kosobuckigordc: downgraded is hypothetical state, what I am trying to do in Openstack is http://docs.aws.amazon.com/AutoScaling/latest/DeveloperGuide/healthcheck.html18:20
marcin_kosobuckischedule bad VM for termination and rebuilt18:21
marcin_kosobuckiin heat room there is no help18:21
gordc"schedule bad VM for termination and rebuilt" would be something that you should ask in heat channel... or on openstack-operators mailing list18:24
marcin_kosobuckiok thank you gordc18:25
gordci don't think we have full equivalence to AWS there. in Aodh, we just added support to track events and you could probably detect if something was 'downgraded'.18:26
*** ildikov has quit IRC18:40
*** julim has quit IRC18:41
*** julim has joined #openstack-ceilometer18:43
*** openstackgerrit has quit IRC18:46
*** openstackgerrit has joined #openstack-ceilometer18:46
*** david-lyle has quit IRC18:55
*** david-lyle has joined #openstack-ceilometer18:55
*** ankita_wagh has quit IRC18:56
*** ankita_wagh has joined #openstack-ceilometer18:59
*** nadya has quit IRC19:04
*** ViswaV has quit IRC19:12
*** _cjones_ has quit IRC19:17
alejandritogordc, jd__ this one https://review.openstack.org/#/c/229461/  doesnt work, because that one sends the policies metric dict with nothing, this line stacks with KeyError https://github.com/openstack/ceilometer/blob/master/ceilometer/dispatcher/gnocchi.py#L34319:21
alejandritogordc, jd__ because it expects a policy on that dict19:21
*** david-lyle has quit IRC19:22
jd__alejandrito: right :/19:22
*** david-lyle has joined #openstack-ceilometer19:22
jd__alejandrito: add a not and -1 on the review I'll fix19:23
gordccrap. forgot how to debug gabbi again.19:25
jd__alejandrito: updated19:29
openstackgerritJulien Danjou proposed openstack/ceilometer: gnocchi: set the default archive policy to None  https://review.openstack.org/22946119:29
jd__gordc: verbose: True in your rule if you want19:29
jd__49926/50000, total: 311.98 seconds, rate: 160.03 done/second -> my initial Gnocchi benchmark on a small VM, creating metrics pretty good19:29
gordcjd__: is there a way to print entire response?19:30
jd__gordc: verbose should do that19:31
openstackgerritPradeep Kilambi proposed openstack/python-gnocchiclient: Add capabilities support to cli  https://review.openstack.org/22958419:31
gordcjd__: got it. thanks19:32
*** david-lyle has quit IRC19:35
*** david-lyle has joined #openstack-ceilometer19:52
*** rakhi has joined #openstack-ceilometer19:57
alejandritojd__, testing20:06
*** david-ly_ has joined #openstack-ceilometer20:07
*** david-lyle has quit IRC20:07
*** david-ly_ is now known as david-lyle20:08
*** sergio_nubeliu has quit IRC20:12
*** shardy has quit IRC20:14
alejandritojd__, seems to work perfect20:19
*** ViswaV has joined #openstack-ceilometer20:20
openstackgerritgordon chung proposed openstack/gnocchi: match archive policy rule based on longest match  https://review.openstack.org/22960020:21
*** ViswaV has quit IRC20:24
*** ViswaV has joined #openstack-ceilometer20:25
*** thumpba has joined #openstack-ceilometer20:25
*** rakhi has quit IRC20:27
*** _cjones_ has joined #openstack-ceilometer20:37
*** adurbin__ has quit IRC20:41
*** _cjones_ has quit IRC20:48
*** MVenesio has quit IRC20:48
*** _cjones_ has joined #openstack-ceilometer20:49
*** ankita_wagh has quit IRC20:52
*** ankita_wagh has joined #openstack-ceilometer20:56
*** thumpba has quit IRC21:01
*** ankita_w_ has joined #openstack-ceilometer21:03
*** ankita_wagh has quit IRC21:03
*** david-lyle has quit IRC21:03
*** david-ly_ has joined #openstack-ceilometer21:03
*** thumpba has joined #openstack-ceilometer21:03
*** david-ly_ is now known as david-lyle21:05
*** thumpba has quit IRC21:08
*** thumpba has joined #openstack-ceilometer21:09
*** thorst has quit IRC21:12
*** david-lyle has quit IRC21:26
*** david-lyle has joined #openstack-ceilometer21:29
*** thumpba_ has joined #openstack-ceilometer21:40
*** Divining has joined #openstack-ceilometer21:41
*** thumpba has quit IRC21:43
*** thumpba_ has quit IRC21:46
*** thumpba has joined #openstack-ceilometer21:46
*** david-lyle has quit IRC21:48
*** ViswaV has quit IRC21:52
*** david-lyle has joined #openstack-ceilometer21:54
*** ViswaV has joined #openstack-ceilometer21:56
openstackgerritgordon chung proposed openstack/gnocchi: match archive policy rule based on longest match  https://review.openstack.org/22960021:58
*** ViswaV_ has joined #openstack-ceilometer21:58
*** gordc has quit IRC21:58
*** ViswaV has quit IRC22:00
*** pradk has quit IRC22:02
openstackgerritRohit Jaiswal proposed openstack/ceilometer: Fix for resource polling warnings  https://review.openstack.org/22965722:02
*** thumpba has quit IRC22:05
*** jruano has quit IRC22:10
*** idegtiarov has quit IRC22:11
*** david-lyle has quit IRC22:15
*** dan-t has quit IRC22:20
*** ddieterly has quit IRC22:28
*** devlaps has quit IRC22:30
*** ankita_wagh has joined #openstack-ceilometer22:30
*** ankita_w_ has quit IRC22:30
*** david-lyle has joined #openstack-ceilometer22:36
*** alejandrito has quit IRC23:04
*** jruano has joined #openstack-ceilometer23:08
*** thorst has joined #openstack-ceilometer23:23
*** thorst has quit IRC23:33
*** jwcroppe has quit IRC23:40
*** jwcroppe has joined #openstack-ceilometer23:41
*** jwcroppe has quit IRC23:45
*** ddieterly has joined #openstack-ceilometer23:49

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