| *** rbak has quit IRC | 00:01 | |
| *** shakamunyi has joined #openstack-ceilometer | 00:02 | |
| *** ryanpetrello has joined #openstack-ceilometer | 00:08 | |
| *** shakamunyi has quit IRC | 00:09 | |
| *** shakamunyi has joined #openstack-ceilometer | 00:10 | |
| *** shakamunyi has quit IRC | 00:15 | |
| *** ryanpetrello has quit IRC | 00:23 | |
| *** Viswanath has joined #openstack-ceilometer | 00:25 | |
| *** Viswanath has quit IRC | 00:29 | |
| *** shakamunyi has joined #openstack-ceilometer | 00:31 | |
| *** changbl has joined #openstack-ceilometer | 00:45 | |
| *** ryanpetrello has joined #openstack-ceilometer | 00:56 | |
| *** yatin has joined #openstack-ceilometer | 01:27 | |
| *** ddieterly has joined #openstack-ceilometer | 01:27 | |
| *** ryanpetrello has quit IRC | 01:39 | |
| *** yatin has quit IRC | 01:40 | |
| *** Viswanath has joined #openstack-ceilometer | 01:47 | |
| *** Viswanath has quit IRC | 01:51 | |
| *** Yanyanhu has joined #openstack-ceilometer | 01:53 | |
| *** ssakhamuri has joined #openstack-ceilometer | 01:56 | |
| *** ssakhamuri has quit IRC | 01:57 | |
| *** nosnos has joined #openstack-ceilometer | 02:06 | |
| *** openstackgerrit has quit IRC | 02:34 | |
| *** ddieterly has quit IRC | 03:00 | |
| *** Viswanath has joined #openstack-ceilometer | 03:22 | |
| *** Viswanath has quit IRC | 03:25 | |
| *** liusheng has quit IRC | 03:31 | |
| *** liusheng has joined #openstack-ceilometer | 03:31 | |
| *** shakamunyi has quit IRC | 03:32 | |
| *** deepthi has joined #openstack-ceilometer | 03:37 | |
| *** boris-42 has quit IRC | 03:37 | |
| *** ddieterly has joined #openstack-ceilometer | 03:51 | |
| *** ddieterly has quit IRC | 03:55 | |
| *** ddieterly has joined #openstack-ceilometer | 03:59 | |
| *** ddieterly has quit IRC | 04:04 | |
| *** liusheng has quit IRC | 04:19 | |
| *** liusheng has joined #openstack-ceilometer | 04:20 | |
| *** yatin has joined #openstack-ceilometer | 04:30 | |
| *** cmyster has quit IRC | 04:39 | |
| *** ddieterly has joined #openstack-ceilometer | 05:00 | |
| *** ddieterly has quit IRC | 05:05 | |
| *** ishant has joined #openstack-ceilometer | 05:40 | |
| *** cmyster has joined #openstack-ceilometer | 05:43 | |
| *** ildikov has quit IRC | 05:58 | |
| *** ddieterly has joined #openstack-ceilometer | 06:00 | |
| *** Yanyanhu has quit IRC | 06:04 | |
| *** ddieterly has quit IRC | 06:05 | |
| *** Yanyanhu has joined #openstack-ceilometer | 06:10 | |
| *** _elena_ has quit IRC | 06:10 | |
| *** _elena_ has joined #openstack-ceilometer | 06:10 | |
| *** boris-42 has joined #openstack-ceilometer | 06:21 | |
| *** k4n0 has joined #openstack-ceilometer | 06:37 | |
| *** _nadya_ has joined #openstack-ceilometer | 06:46 | |
| *** nellysmitt has joined #openstack-ceilometer | 06:56 | |
| *** ddieterly has joined #openstack-ceilometer | 07:00 | |
| *** ddieterly has quit IRC | 07:01 | |
| *** ddieterl_ has joined #openstack-ceilometer | 07:01 | |
| *** eglynn_ has joined #openstack-ceilometer | 07:01 | |
| *** ddieterl_ has quit IRC | 07:06 | |
| *** nellysmitt has quit IRC | 07:14 | |
| *** cmyster has quit IRC | 07:20 | |
| *** ildikov has joined #openstack-ceilometer | 07:31 | |
| *** nellysmitt has joined #openstack-ceilometer | 07:36 | |
| *** cmyster has joined #openstack-ceilometer | 07:39 | |
| *** Ala has joined #openstack-ceilometer | 07:46 | |
| *** _nadya_ has quit IRC | 07:50 | |
| *** Longgeek has joined #openstack-ceilometer | 08:04 | |
| *** Longgeek has quit IRC | 08:06 | |
| *** Longgeek has joined #openstack-ceilometer | 08:07 | |
| *** ishant2 has joined #openstack-ceilometer | 08:08 | |
| *** ishant has quit IRC | 08:10 | |
| *** ildikov has quit IRC | 08:13 | |
| *** nellysmitt has quit IRC | 08:13 | |
| *** promulo has joined #openstack-ceilometer | 08:24 | |
| *** ildikov has joined #openstack-ceilometer | 08:25 | |
| *** nellysmitt has joined #openstack-ceilometer | 08:32 | |
| *** ifarkas has joined #openstack-ceilometer | 08:41 | |
| *** safchain has joined #openstack-ceilometer | 08:47 | |
| *** ana_ has joined #openstack-ceilometer | 09:09 | |
| *** ana_ is now known as malagon | 09:09 | |
| *** nellysmitt has quit IRC | 09:17 | |
| *** nellysmitt has joined #openstack-ceilometer | 09:19 | |
| *** nellysmitt has quit IRC | 09:22 | |
| *** moravec has quit IRC | 09:24 | |
| *** malagon has quit IRC | 09:27 | |
| *** Yanyanhu has quit IRC | 09:41 | |
| *** amalagon has joined #openstack-ceilometer | 09:49 | |
| *** IvanBerezovskiy has quit IRC | 09:51 | |
| *** IvanBerezovskiy has joined #openstack-ceilometer | 09:51 | |
| *** cdent has joined #openstack-ceilometer | 09:55 | |
| *** _nadya_ has joined #openstack-ceilometer | 09:57 | |
| *** _nadya_ has quit IRC | 10:06 | |
| *** Longgeek has quit IRC | 10:09 | |
| *** Longgeek has joined #openstack-ceilometer | 10:10 | |
| *** Longgeek_ has joined #openstack-ceilometer | 10:12 | |
| *** Longgeek has quit IRC | 10:15 | |
| *** eoutin has joined #openstack-ceilometer | 10:29 | |
| *** cdent has quit IRC | 10:30 | |
| *** eoutin has quit IRC | 10:36 | |
| *** nosnos has quit IRC | 10:39 | |
| *** eoutin has joined #openstack-ceilometer | 10:50 | |
| *** ishant2 has quit IRC | 10:52 | |
| *** ishant has joined #openstack-ceilometer | 10:53 | |
| *** eoutin has quit IRC | 10:55 | |
| *** cdent has joined #openstack-ceilometer | 10:56 | |
| *** nosnos has joined #openstack-ceilometer | 11:02 | |
| *** nsaje has quit IRC | 11:19 | |
| *** deepthi has quit IRC | 11:30 | |
| *** ishant has quit IRC | 11:38 | |
| *** ishant has joined #openstack-ceilometer | 11:39 | |
| *** nellysmitt has joined #openstack-ceilometer | 11:53 | |
| *** chmouel has joined #openstack-ceilometer | 11:59 | |
| chmouel | hello there i am having an issue with pecan and ceilometer gate job here, http://logs.openstack.org/10/131410/2/gate/gate-pecan-tox-ceilometer-tip/b119a86/console.html | 12:00 |
|---|---|---|
| chmouel | it seems that pymemcache is broken there | 12:00 |
| chmouel | it has been happening since Friday on that pecan job, did you guys have any issues as well? | 12:00 |
| chmouel | uit | 12:01 |
| *** ildikov has quit IRC | 12:04 | |
| cdent | chmouel: I haven't heard of anything recently, but I may be out of touch | 12:08 |
| cdent | pymemcache is brought in from tooz | 12:09 |
| cdent | and there was a new release 7 days ago | 12:10 |
| cdent | ah, I see the problem | 12:12 |
| cdent | chmouel: the wrong package is being downloaded pymemcache-1.2.6.macosx-10.9-intel.tar.gz | 12:12 |
| *** yatin has quit IRC | 12:13 | |
| cdent | it's unclear why pip is selecting that version | 12:14 |
| cdent | If tooz is installed in a virtenv directly, pymemcache installs fine | 12:16 |
| cdent | when trying to install ceilometer tip from git, things fail (as the logs show) | 12:16 |
| *** asalkeld has left #openstack-ceilometer | 12:17 | |
| *** ildikov has joined #openstack-ceilometer | 12:19 | |
| chmouel | ah thanks cdent this is weird indeed, | 12:20 |
| cdent | I posted the above on the patchset too, for reference | 12:21 |
| cdent | I think it is just that the mac binary release is not playing well with... something? | 12:21 |
| chmouel | it's very weird pip is trying to install the macosx binary at first | 12:22 |
| cdent | yes | 12:22 |
| cdent | what I don't understand is why that weirdness only kicks in when installed ceilo from git | 12:23 |
| chmouel | let's see if we can ping the pip maints i think the hangout on infra | 12:24 |
| *** exploreshaifali has joined #openstack-ceilometer | 12:37 | |
| *** ryanpetrello has joined #openstack-ceilometer | 12:38 | |
| *** ddieterly has joined #openstack-ceilometer | 12:46 | |
| *** ildikov has quit IRC | 12:48 | |
| *** cmyster has quit IRC | 12:48 | |
| *** ildikov has joined #openstack-ceilometer | 12:58 | |
| *** zqfan has joined #openstack-ceilometer | 12:59 | |
| *** nosnos has quit IRC | 13:10 | |
| *** Longgeek has joined #openstack-ceilometer | 13:33 | |
| *** Longgeek_ has quit IRC | 13:36 | |
| *** amalagon has quit IRC | 13:37 | |
| *** nellysmitt has quit IRC | 13:41 | |
| *** _elena_ has quit IRC | 13:41 | |
| *** ddieterly has quit IRC | 13:43 | |
| *** yassine has joined #openstack-ceilometer | 13:45 | |
| *** k4n0 has quit IRC | 13:46 | |
| *** ishant has quit IRC | 14:01 | |
| *** _nadya__ has joined #openstack-ceilometer | 14:04 | |
| *** julim has joined #openstack-ceilometer | 14:07 | |
| *** joesavak has joined #openstack-ceilometer | 14:07 | |
| *** _elena_ has joined #openstack-ceilometer | 14:10 | |
| *** fnaval has joined #openstack-ceilometer | 14:12 | |
| *** IvanBerezovskiy has left #openstack-ceilometer | 14:13 | |
| *** _nadya__ has quit IRC | 14:16 | |
| *** _nadya_ has joined #openstack-ceilometer | 14:16 | |
| *** ddieterly has joined #openstack-ceilometer | 14:20 | |
| *** cmyster has joined #openstack-ceilometer | 14:26 | |
| *** cmyster has joined #openstack-ceilometer | 14:26 | |
| *** fnaval has quit IRC | 14:26 | |
| *** shakamunyi has joined #openstack-ceilometer | 14:31 | |
| *** ana_ has joined #openstack-ceilometer | 14:31 | |
| *** thomasem has joined #openstack-ceilometer | 14:33 | |
| *** amalagon has joined #openstack-ceilometer | 14:35 | |
| *** ana_ has quit IRC | 14:36 | |
| *** adam_g` is now known as adam_g | 14:37 | |
| *** adam_g has quit IRC | 14:37 | |
| *** adam_g has joined #openstack-ceilometer | 14:37 | |
| *** exploreshaifali has quit IRC | 14:39 | |
| *** openstackgerrit has joined #openstack-ceilometer | 14:40 | |
| openstackgerrit | Lena Novokshonova proposed openstack/ceilometer: Add new notifications types for volumes/snapshots https://review.openstack.org/131147 | 14:51 |
| eglynn_ | jd__: once https://review.openstack.org/132092 lands, can we get a tooz 0.9.0 released? | 14:53 |
| jd__ | eglynn_: sure | 14:53 |
| eglynn_ | jd__: excellent, thank you sir! :) | 14:54 |
| openstackgerrit | Julien Danjou proposed stackforge/gnocchi: storage: multi-thread add_measure in Carbonara based drivers https://review.openstack.org/132681 | 14:58 |
| openstackgerrit | Julien Danjou proposed stackforge/gnocchi: storage: factorize carbonara based drivers https://review.openstack.org/133973 | 14:58 |
| *** ildikov has quit IRC | 15:10 | |
| *** underyx is now known as underyx|off | 15:16 | |
| *** nsaje has joined #openstack-ceilometer | 15:19 | |
| *** ryanpetrello has quit IRC | 15:19 | |
| *** rbak has joined #openstack-ceilometer | 15:25 | |
| *** ryanpetrello has joined #openstack-ceilometer | 15:25 | |
| *** Ala has quit IRC | 15:27 | |
| *** ildikov has joined #openstack-ceilometer | 15:30 | |
| *** Longgeek has quit IRC | 15:35 | |
| *** _nadya_ has quit IRC | 15:38 | |
| *** cmyster has quit IRC | 15:44 | |
| eglynn_ | jd__: two quick questions about gnocchi if you got a sec? | 15:45 |
| eglynn_ | first (IIRC this was mentioned at summit) would it be good to allow a "full-res" granularity (e.g. 0s or -1) to be specified in the archive policy? | 15:46 |
| eglynn_ | (in addition to the "green bar" representing in the ingestion buffer / look-back window) | 15:46 |
| eglynn_ | this would allow alarming on a entity to have *any* reasonable period, as long as the number of eval periods is relatively small | 15:46 |
| eglynn_ | (so that the span of the GET .../measures query fits within the retention of the full-res timeseries, and aggregation could be done on-demand) | 15:47 |
| eglynn_ | second, any ideas on how we could represent transient resource-to-resource associations in the brave new world of gnocchi? | 15:49 |
| eglynn_ | (i.e. without any per-datapoint metadata or tags) | 15:49 |
| * eglynn_ is thinking here of opencontrail network flows, where each measure is associated with a "here" network and a "there" network | 15:50 | |
| boris-42 | eglynn_: h there | 15:52 |
| eglynn_ | boris-42: hey | 15:52 |
| boris-42 | eglynn_: so I replied on your comment | 15:53 |
| boris-42 | eglynn_: if you have any questions about voting stuff just ping me | 15:53 |
| eglynn_ | boris-42: looking | 15:53 |
| jd__ | eglynn_: so, about 1st question, yes, that'd be the back window parameter (not yet exposed in the API, but soon) and ask for non-aggregated point (recent patch implementing that with full=yes in carbonara and the drivers) | 15:58 |
| *** cmyster has joined #openstack-ceilometer | 15:59 | |
| jd__ | eglynn_: for your 2nd question, I'd build a resource type "network" and have the entity associated to a network with the other network as the key/name | 15:59 |
| *** cmyster has quit IRC | 15:59 | |
| *** cmyster has joined #openstack-ceilometer | 16:00 | |
| *** cmyster has joined #openstack-ceilometer | 16:00 | |
| eglynn_ | jd__: on the 1st, I was thinking more of the query asking for say period=120 when the archive policy granularity is say [1min, 1hour] | 16:02 |
| eglynn_ | ... in that case gnocchi could return the requested 120s periodization by aggregating the full-res data on the fly | 16:02 |
| eglynn_ | (if the timespan of the query was short and recent enough) | 16:02 |
| *** david-lyle has joined #openstack-ceilometer | 16:03 | |
| eglynn_ | i.e. not returning the non-aggregated data, instead revert to the classic ceilo aggregate-on-demand pattern if pre-aggregated values not available for that granularity | 16:03 |
| jd__ | eglynn_: ok I totally misunderstood your question then | 16:04 |
| jd__ | that'd be good enough for me, though I don't know if it's something we should implement in the API or in the storage driver | 16:05 |
| jd__ | I wonder if there's any driver that could implement that on storage side | 16:05 |
| eglynn_ | jd__: fair enough ... I was thinking exposing in the API might be nice as the ingestion buffer is kinda an artifact of the implementation (IIUC) | 16:06 |
| jd__ | that's not related to the buffer anyway | 16:07 |
| jd__ | IIUC | 16:07 |
| jd__ | unless you want to have that period computed only on it? | 16:07 |
| jd__ | that'd be a very edgy case | 16:08 |
| jd__ | all drivers have an ingestion buffer, it's just that it's easier to have access to the one from Carbonara from other I imagine | 16:08 |
| eglynn_ | jd__: yeah, I was thinking having the periodization computed on the full-res buffer *if* the requested granularity wasn't being pre-aggregated for that entity | 16:10 |
| eglynn_ | jd__: e.g. if the requested period was say 120s, then the pre-aggregated periods for this entity were say 1min and 1hour | 16:10 |
| eglynn_ | jd__: just to clarify the 2nd q, did you mean exposing *two* levels of naming in the resource-to-entity mapping? | 16:12 |
| eglynn_ | e.g. ... GET /v1/resource/network/here_UUID/entity/in.bytes/there_UUID/measures | 16:12 |
| jd__ | eglynn_: the problem is that full-res buffer is only 1 hour long in that case, and I don't know if we have access to it in other drivers | 16:14 |
| jd__ | eglynn_: 2nd q: no I meant not changing anything just /v1/resources/network/here_uuid/entity/there_uuid.in.bytes/measures | 16:15 |
| eglynn_ | jd__: yeah, so in the alarming case, that 1 hour might often be enough as the period*eval_periods of the alarm is generally short | 16:17 |
| eglynn_ | jd__: ... but if we allowed the archive policy to explicitly allow for a longer full-res rentention, that might be more logicial for users | 16:17 |
| jd__ | eglynn_: I'm totally with all of that if all the drivers can have this ability | 16:18 |
| jd__ | eglynn_: so let me know if we can do that with Influx :p | 16:18 |
| eglynn_ | e.g. archive_policy = [(granularity=0s, retention=2hrs), (granularity=1min, retention=1day), (granularity=1hr, retention=1week)] | 16:18 |
| *** renatoarmani has joined #openstack-ceilometer | 16:18 | |
| eglynn_ | where 0s ==full-res | 16:18 |
| eglynn_ | coolness, will do :) | 16:19 |
| jd__ | eglynn_: btw there's some patches to review if you ever feel like it O;-) | 16:19 |
| eglynn_ | jd__: yeah, sorry, I've been a bit remiss on that since summit ... will give the gnocchi queue on gerrit another pass before EoD | 16:20 |
| jd__ | cool thanks | 16:21 |
| *** _cjones_ has joined #openstack-ceilometer | 16:34 | |
| *** Longgeek has joined #openstack-ceilometer | 16:35 | |
| *** Viswanath has joined #openstack-ceilometer | 16:36 | |
| *** _nadya_ has joined #openstack-ceilometer | 16:36 | |
| *** Viswanath has quit IRC | 16:39 | |
| *** Longgeek has quit IRC | 16:40 | |
| boris-42 | eglynn_: can we discuss rally job? | 16:42 |
| boris-42 | eglynn_: if you have some free slots of time=) | 16:42 |
| boris-42 | eglynn_: cause you are missing the point.. | 16:42 |
| boris-42 | a bit=) | 16:42 |
| eglynn_ | boris-42: missing the point with my latest comment on https://review.openstack.org/#/c/129922/7/specs/kilo/rally-check-gate.rst ? | 16:42 |
| boris-42 | eglynn_: ya | 16:43 |
| eglynn_ | boris-42: sure, always open to correction ... shoot | 16:43 |
| boris-42 | eglynn_: I just read it | 16:43 |
| boris-42 | eglynn_: so you are absolutelly right about difference between nodes | 16:43 |
| boris-42 | eglynn_: and perf of them | 16:43 |
| eglynn_ | boris-42: OK, so in this case we avoid that reflecting in rally timings how? | 16:44 |
| boris-42 | eglynn_: so the idea of sla checks (that are related to duration) can be something like (value) that is bigger then in case of 99% | 16:44 |
| boris-42 | eglynn_: then you get failed rally job in 1% cases | 16:44 |
| boris-42 | eglynn_: that's all | 16:44 |
| *** renatoarmani has quit IRC | 16:45 | |
| eglynn_ | boris-42: "bigger then in case of 99%" == 99th percentile of previous/recent test runs? | 16:45 |
| boris-42 | eglynn_: nope | 16:45 |
| boris-42 | eglynn_: just do N times rechecks and take a look at what avarage is | 16:45 |
| boris-42 | eglynn_: it will require some amount of time to collect data (so at first point we can just skip this step) | 16:46 |
| boris-42 | eglynn_: and when we collect enough info we will just put some criterias of success | 16:46 |
| boris-42 | eglynn_: this really can help you to avoid bad patches, that have huge impact on performance. | 16:46 |
| eglynn_ | boris-42: do you mean N manual rechecks triggered by owner of a single patch, or some auto-rechecking? | 16:46 |
| amalagon | hi jd__: has anything changed in the way gnocchi is built? I am trying to add an entry point in setup.cfg but even though I'm following the same steps I've done before, stevedore can't find it | 16:47 |
| boris-42 | eglynn_: we can just add job + some benchmarks | 16:47 |
| boris-42 | eglynn_: then adjust numbers | 16:47 |
| dhellmann | amalagon: https://pypi.python.org/pypi/entry_point_inspector might help debug | 16:47 |
| boris-42 | eglynn_: and when it will be some amount of Numbers collect all json files from gates and find these values | 16:47 |
| eglynn_ | amalagon: have you re-run the setup.py | 16:47 |
| eglynn_ | ? | 16:47 |
| boris-42 | eglynn_: so we won't need to do anything by hand | 16:47 |
| amalagon | hey eglynn_, yeah | 16:47 |
| amalagon | dhellmann: thanks, I'll try that | 16:48 |
| eglynn_ | amalagon: k, in that case wot dhellmann said | 16:48 |
| boris-42 | eglynn_: and about historical perfromance data. We will just provide a page on github.io | 16:48 |
| boris-42 | eglynn_: with results for major components | 16:49 |
| jd__ | amalagon: did you run setup.py? | 16:49 |
| boris-42 | eglynn_: so this is just used for tracking trends (but not for catching bad patches) | 16:49 |
| boris-42 | that will be used* | 16:49 |
| eglynn_ | boris-42: OK, just so I understand, the collection and aggregation of the historical data is automated? | 16:49 |
| boris-42 | eglynn_: nope | 16:49 |
| amalagon | jd__: yep, if 'sudo python setup.py install' is right | 16:49 |
| boris-42 | eglynn_: it will be automated. But for regression testing you don't need that | 16:50 |
| jd__ | amalagon: should be yeah | 16:50 |
| jd__ | amalagon: you could try to debug with entry_point_inspector | 16:50 |
| boris-42 | eglynn_: just put criteria of success like no failures + some amount of stuff | 16:50 |
| amalagon | jd__: :) thanks, gonna try that now | 16:50 |
| boris-42 | eglynn_: hm maybe if we make hangout call it will be simpler for you to understand=) | 16:50 |
| boris-42 | eglynn_: if I just show | 16:50 |
| boris-42 | =) | 16:50 |
| eglynn_ | boris-42: the point is that setting that regression criteria is not trivial if we want to avoid false positives and yet not set it too loosely that it never catches a real performance issue | 16:52 |
| boris-42 | eglynn_: it's not that hard as you think | 16:52 |
| boris-42 | eglynn_: let's add propers jobs without criterias (except no failures) | 16:52 |
| boris-42 | eglynn_: and we will just incremental work on regression criterias | 16:53 |
| *** shakamunyi has quit IRC | 16:53 | |
| boris-42 | eglynn_: it is really hard at one step to do, but it really should be our long term goal | 16:53 |
| eglynn_ | boris-42: can you define "incremental work on regression criteria" if no aggregation of the data? | 16:53 |
| boris-42 | eglynn_: every run => new results | 16:54 |
| eglynn_ | boris-42: do you mean manually gather the test results from the CI nodes? | 16:54 |
| boris-42 | eglynn_: why manually just make small script | 16:54 |
| boris-42 | eglynn_: that will fetch json files, and do analyze that we need | 16:54 |
| eglynn_ | boris-42: is that the approach used for the rally SLA tests in other projects? | 16:54 |
| *** ifarkas has quit IRC | 16:54 | |
| boris-42 | eglynn_: not yet | 16:54 |
| boris-42 | eglynn_: you can be the first one lol) | 16:55 |
| openstackgerrit | ZhiQiang Fan proposed openstack/python-ceilometerclient: Add --slowest option for testr https://review.openstack.org/134003 | 16:56 |
| boris-42 | eglynn_: I mean json ouput of rally is quite simple | 16:56 |
| eglynn_ | boris-42: k, so that was the point that I missed? | 16:56 |
| boris-42 | eglynn_: so we will just fetch all JSON files from all check jobs that passed other unit tess | 16:56 |
| boris-42 | eglynn_: and functional | 16:57 |
| boris-42 | eglynn_: and analyze duration and find just the magic number | 16:57 |
| eglynn_ | boris-42: i.e. that it's possible to manually aggregate the test results, figure out an average over a period, add in some wiggle-room, and then bada-bing ... we're done! | 16:57 |
| amalagon | dhellmann: hi, sorry, basic question here- if I see my desired entry point listed in gnocchi.egg-info/entry_points.txt but epi group list doesn't show my entry point, does that tell me something useful about what's going on? | 16:58 |
| boris-42 | eglynn_: ya | 16:58 |
| boris-42 | eglynn_: but we need some amount of Runs for that | 16:58 |
| boris-42 | eglynn_: so at the start we won't have any criteria (except no failures) | 16:59 |
| *** ildikov has quit IRC | 16:59 | |
| boris-42 | eglynn_: and then we will work on automation of adjusting numbers | 16:59 |
| boris-42 | eglynn_: from results from gates | 16:59 |
| boris-42 | eglynn_: and we will get some kind of regressions tests | 16:59 |
| eglynn_ | boris-42: yeah, sure I get that it can be done ... I suspect it's not quite as trivial as you present | 17:01 |
| eglynn_ | boris-42: for example we'd have to look into the data and conisder excluding outliers | 17:02 |
| eglynn_ | boris-42: it would be interesting also to see if those numbers revealed any sustained differences between the RAX and HP public clouds | 17:02 |
| eglynn_ | (since a CI node can be scheduled to either IIUC) | 17:02 |
| boris-42 | eglynn_: so from what I saw the difference is < 2 times | 17:03 |
| *** joesavak has quit IRC | 17:03 | |
| boris-42 | eglynn_: but what I think is we need just to start working on this=) | 17:04 |
| *** joesavak has joined #openstack-ceilometer | 17:04 | |
| eglynn_ | boris-42: "difference is < 2 times" ==> instances on one cloud up to twice as fast as the other cloud? | 17:05 |
| boris-42 | eglynn_: ya | 17:06 |
| boris-42 | eglynn_: that is what I saw in rally results for a quite long amount of time | 17:06 |
| eglynn_ | boris-42: ... so in that case, surely we'd need cloud-specific thresholds for the SLAs? | 17:06 |
| boris-42 | eglynn_: I saw a lot of regressions that >2 times | 17:07 |
| *** yassine has quit IRC | 17:07 | |
| eglynn_ | boris-42: i.e. expect number < X on HP, but < Y on RAX | 17:07 |
| *** exploreshaifali has joined #openstack-ceilometer | 17:08 | |
| boris-42 | eglynn_: so what I think is for the begging it's nice to get just absolute values | 17:08 |
| boris-42 | eglynn_: at least* | 17:08 |
| boris-42 | eglynn_: you can't do everything at one step+) | 17:09 |
| eglynn_ | boris-42: so the absolute value should be based on the aggregated numbers from the cloud that's consistently slowest, amight? | 17:09 |
| eglynn_ | boris-42: ... i.e. we discard the historical numbers from the faster cloud | 17:09 |
| *** zqfan has quit IRC | 17:10 | |
| boris-42 | eglynn_: I will just drop this stuff (about different clouds) and found value that has passed more then X% times | 17:10 |
| boris-42 | eglynn_: like 99% or 99.9% or any | 17:10 |
| boris-42 | eglynn_: this will make job quite rare failing and quite tight to what we can do for now | 17:11 |
| boris-42 | eglynn_: and we will still work on "normalization" of results | 17:12 |
| boris-42 | eglynn_: like run small set of benchmark and get some magic coefficient* | 17:12 |
| boris-42 | eglynn_: what we need actually is to measure perfromance of memory, cpu, disks, mysql | 17:13 |
| boris-42 | eglynn_: and make formula using this coefficients=) | 17:13 |
| eglynn_ | boris-42: "found value that has passed more then X% times like 99%" is the 99th percentile that I referred to above, IIUC | 17:15 |
| *** nellysmitt has joined #openstack-ceilometer | 17:17 | |
| eglynn_ | boris-42: if there is such a sustained difference between the numbers from the two clouds, aggregating over both is going to be less meaningful | 17:18 |
| *** cdent has quit IRC | 17:18 | |
| boris-42 | eglynn_: so we can think about this a bit, but I still think that normalization is the better way.. | 17:19 |
| boris-42 | eglynn_: then have different SLA for different clouds.. | 17:20 |
| boris-42 | eglynn_: in any case I don't see anything that should block this work=) | 17:20 |
| eglynn_ | boris-42: a-ha, I see ... normalization == "multiply the numbers from one cloud by a compensating factor" ? | 17:21 |
| boris-42 | eglynn_: yep | 17:21 |
| eglynn_ | boris-42: fair enough, if the penalty for using one cloud over the other is relatively constant | 17:22 |
| boris-42 | eglynn_: and factor is calculated from formula that includes (cpu, mem, disk) perfomrace | 17:22 |
| boris-42 | eglynn_: factor can be every time calculated | 17:22 |
| boris-42 | eglynn_: running small benchmarks inside vm | 17:22 |
| boris-42 | eglynn_: before task | 17:22 |
| *** _nadya_ has quit IRC | 17:22 | |
| boris-42 | eglynn_: so this is very interesting topic for PHD =) | 17:23 |
| eglynn_ | boris-42: ok, so is the kind of info I was looking for ... I'll summarize in a comment on gerrit | 17:24 |
| boris-42 | eglynn_: great | 17:24 |
| dhellmann | amalagon: a couple of things could cause that. are epi and gnocchi installed into the same site-packages (globally or in a virtualenv)? | 17:28 |
| amalagon | uh, great question - I think they are both installed globally because I did 'pip install entry_point_inspector' and installed gnocchi globally, but I'll double-check the site-packages | 17:30 |
| eglynn_ | boris-42: done on https://review.openstack.org/#/c/129922/7/specs/kilo/rally-check-gate.rst | 17:33 |
| amalagon | dhellmann: ok, yep, they are both installed in user/local/lib/python2.7/dist-packages | 17:33 |
| *** renatoarmani has joined #openstack-ceilometer | 17:34 | |
| dhellmann | amalagon: so when you look at the entry_points.txt file under the gnocchi installed in dist-packages you see the entry point, but when you scan with epi you don't see it? | 17:35 |
| *** _nadya_ has joined #openstack-ceilometer | 17:35 | |
| amalagon | dhellmann: exactly; also, there are two other entry points in the setup.cfg - both of these appear with epi, just not the additional one I made. | 17:37 |
| dhellmann | amalagon: how did you get the new entry point listed in the entry_points.txt file? | 17:37 |
| amalagon | dhellmann: sorry, don't understand the question? | 17:38 |
| dhellmann | amalagon: at one point there was no new entry point and now there is. What happened between those 2 states to cause it to be added to the file? Did you edit entry_points.txt, setup.cfg, something else? | 17:38 |
| *** fnaval has joined #openstack-ceilometer | 17:39 | |
| *** _nadya_ has quit IRC | 17:39 | |
| amalagon | ah, I see. So I added my entry point into setup.cfg | 17:40 |
| dhellmann | ok, that wouldn't do it by itself, so did you re-install? | 17:41 |
| amalagon | yep, and then did the sudo python setup.py install | 17:41 |
| dhellmann | can you put your setup.cfg and entry_points.txt files on http://paste.openstack.org? | 17:42 |
| amalagon | yep! one sec | 17:42 |
| dhellmann | amalagon: also the output of epi | 17:42 |
| amalagon | my setup.cfg file: http://paste.openstack.org/show/132448 , entry_points.txt: http://paste.openstack.org/show/132449 , results of epi: http://paste.openstack.org/show/132450 | 17:45 |
| dhellmann | amalagon: and which plugin is the new one? | 17:45 |
| amalagon | dhellmann: gnocchi.aggregate | 17:46 |
| dhellmann | amalagon: ok, that's a namespace not an individual plugin | 17:47 |
| dhellmann | amalagon: what does "epi group show gnocchi.aggregate" report? | 17:47 |
| amalagon | dhellmann: http://paste.openstack.org/show/132451 | 17:48 |
| dhellmann | amalagon: what does this print?: python -c 'import gnocchi; print gnocchi.__file__' | 17:49 |
| amalagon | hm, AttributeError: 'module' object has no attribute '__file__' ? | 17:50 |
| dhellmann | replace __file__ with __path__ and try again | 17:50 |
| amalagon | same error, no attribute '__path__' | 17:51 |
| dhellmann | ok, so your gnocchi installation seems broken somehow | 17:51 |
| amalagon | nooo | 17:51 |
| dhellmann | try "pip uninstall gnocchi" and see if that is able to remove anything | 17:51 |
| *** safchain has quit IRC | 17:51 | |
| amalagon | whoa! I uninstalled and reinstalled and the namespace showed up! | 17:52 |
| amalagon | thank you! | 17:52 |
| dhellmann | good! | 17:52 |
| amalagon | :D | 17:53 |
| *** shakamunyi has joined #openstack-ceilometer | 17:56 | |
| *** shakamunyi has quit IRC | 17:56 | |
| *** fnaval has quit IRC | 17:59 | |
| *** shakamunyi has joined #openstack-ceilometer | 18:01 | |
| *** alexpilotti has joined #openstack-ceilometer | 18:21 | |
| *** eglynn_ is now known as eglynn-afk | 18:22 | |
| *** ildikov has joined #openstack-ceilometer | 18:27 | |
| *** renatoarmani has quit IRC | 18:32 | |
| *** renatoarmani has joined #openstack-ceilometer | 18:35 | |
| *** jsavak has joined #openstack-ceilometer | 18:58 | |
| *** joesavak has quit IRC | 19:02 | |
| *** alexpilotti_ has joined #openstack-ceilometer | 19:09 | |
| *** liusheng has quit IRC | 19:09 | |
| *** alexpilotti has quit IRC | 19:09 | |
| *** alexpilotti_ is now known as alexpilotti | 19:09 | |
| *** liusheng has joined #openstack-ceilometer | 19:09 | |
| *** shakamunyi has quit IRC | 19:12 | |
| *** shakamunyi has joined #openstack-ceilometer | 19:13 | |
| *** rbak_ has joined #openstack-ceilometer | 19:44 | |
| *** rbak has quit IRC | 19:44 | |
| *** ddieterly has quit IRC | 19:47 | |
| *** ddieterly has joined #openstack-ceilometer | 19:48 | |
| *** ddieterly has quit IRC | 19:53 | |
| *** ddieterly has joined #openstack-ceilometer | 19:55 | |
| *** dnalezyt has joined #openstack-ceilometer | 19:55 | |
| *** jsavak has quit IRC | 20:01 | |
| *** joesavak has joined #openstack-ceilometer | 20:03 | |
| *** prad has joined #openstack-ceilometer | 20:22 | |
| *** exploreshaifali has quit IRC | 20:24 | |
| *** alexpilotti has quit IRC | 20:33 | |
| *** promulo has quit IRC | 20:34 | |
| *** promulo has joined #openstack-ceilometer | 20:35 | |
| *** _nadya_ has joined #openstack-ceilometer | 20:39 | |
| *** _nadya_ has quit IRC | 20:43 | |
| *** renatoarmani has quit IRC | 20:44 | |
| *** amalagon has quit IRC | 20:46 | |
| *** shakamunyi_ has joined #openstack-ceilometer | 20:49 | |
| *** thomasem has quit IRC | 20:57 | |
| *** shakamunyi_ has quit IRC | 21:03 | |
| *** amalagon has joined #openstack-ceilometer | 21:05 | |
| *** shakamun_ has joined #openstack-ceilometer | 21:10 | |
| *** shakamunyi has quit IRC | 21:10 | |
| *** MasterPiece has joined #openstack-ceilometer | 21:11 | |
| *** shakamunyi_ has joined #openstack-ceilometer | 21:17 | |
| *** shakamun_ has quit IRC | 21:26 | |
| *** shakamunyi_ has quit IRC | 21:27 | |
| *** shakamunyi has joined #openstack-ceilometer | 21:30 | |
| *** Viswanath has joined #openstack-ceilometer | 21:36 | |
| *** thomasem has joined #openstack-ceilometer | 21:37 | |
| *** Viswanath has quit IRC | 21:40 | |
| *** Viswanath has joined #openstack-ceilometer | 21:41 | |
| *** Viswanath has quit IRC | 21:44 | |
| *** asalkeld has joined #openstack-ceilometer | 21:56 | |
| *** joesavak has quit IRC | 22:16 | |
| *** nellysmitt has quit IRC | 22:22 | |
| *** asalkeld has quit IRC | 22:51 | |
| *** thomasem has quit IRC | 22:54 | |
| *** ddieterly has quit IRC | 23:05 | |
| *** _cjones_ has quit IRC | 23:06 | |
| *** asalkeld has joined #openstack-ceilometer | 23:07 | |
| *** _cjones_ has joined #openstack-ceilometer | 23:09 | |
| *** eglynn-afk has quit IRC | 23:09 | |
| *** david-lyle is now known as david-lyle_afk | 23:12 | |
| *** promulo has quit IRC | 23:22 | |
| *** promulo has joined #openstack-ceilometer | 23:22 | |
| *** promulo__ has joined #openstack-ceilometer | 23:28 | |
| *** promulo has quit IRC | 23:29 | |
| *** ryanpetrello has quit IRC | 23:39 | |
| *** prad has quit IRC | 23:42 | |
| *** exploreshaifali has joined #openstack-ceilometer | 23:48 | |
| *** rbak_ has quit IRC | 23:53 | |
| *** eglynn-afk has joined #openstack-ceilometer | 23:57 | |
| *** shakamunyi has quit IRC | 23:58 | |
| *** shakamunyi has joined #openstack-ceilometer | 23:58 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!