Can't read power.typical_10_mobile timeseries with timestamp |
||
Issue description
The following request:
dashboard_services.Timeseries2(
test_suite='power.typical_10_mobile',
bot='ChromiumPerf:android-go-perf',
measurement='application_energy_consumption_mwh',
test_case='http___de.m.wikipedia.org_wiki_K%C3%B6lner_Dom',
columns='revision,avg')
Returns a bunch of expected data. But adding "timestamp" to the list of columns, e.g.:
dashboard_services.Timeseries2(
test_suite='power.typical_10_mobile',
bot='ChromiumPerf:android-go-perf',
measurement='application_energy_consumption_mwh',
test_case='http___de.m.wikipedia.org_wiki_K%C3%B6lner_Dom',
columns='revision,avg,timestamp')
returns:
500 Internal Server Error
The server has either erred or is incapable of performing the requested operation.
Similarly adding a min_timestamp parameter appears to cause the same problem.
I don't have this problem with timeseries from other test suites.
,
Nov 26
Thanks for reporting!
The log says NeedIndexError.
index.yaml contains this index:
- kind: Row
properties:
- name: parent_test
- name: revision
- name: timestamp
- name: value
The log says it tried to use this non-existent index:
- kind: Row
properties:
- name: parent_test
- name: revision
direction: desc
- name: timestamp
- name: value
Adding a Row index is very expensive and probably not necessary here.
Timeseries2Handler._RowQueryProjection is responsible for deciding when to do a projection query. Maybe it should not try to use a projection when timestamp is required, or maybe it should sort the query by revision ascending instead of descending so it uses the index that already exists.
I have no idea why this wouldn't be a problem for other suites. :-|
perezju: Can you help me prioritize this? Is this blocking anything specific?
,
Nov 27
> I have no idea why this wouldn't be a problem for other suites. :-| Yeah, that's really strange. :-| Anyway, at the moment this is only blocking the cleanup in issue 900218, to migrate the last of the studies from timeseries v1 to v2; but I believe that this study can live in v1 for a while longer. I'll drop to P3 for now as that seems more appropriate. |
||
►
Sign in to add a comment |
||
Comment 1 by perezju@chromium.org
, Nov 23