some monarch metrics claim not to have a retention policy |
|||||||
Issue descriptionhttps://viceroy.corp.google.com/chromeos/cl-actions?duration=8d While I know there have been CQ issues lately, I'm pretty sure this data is not correct.
,
Oct 13 2017
Possibly related, this query http://shortn/_haPvfXg7wr reports that the metric is not retained by any chrome-infra retention policy, whereas it certainly used to be a few days ago.
,
Oct 13 2017
This looks to me like either a monarch bug, or some kind of change in retention. The same effective query shows different results based on whether streams are filtered during fetch or after fetch: http://shortn/_2hOcVnlGIY
,
Oct 13 2017
+some monarch advisors
,
Oct 13 2017
+estaab who owns monarch things these days?
,
Oct 13 2017
katthomas might be able to help, but this might be too chromeos-specific.
,
Oct 13 2017
Is chrome-infra-retentions running out of quota?
,
Oct 13 2017
Repository=Force gets rid of this inconsistency, on pcon. However, query still misbehaving in viceroy.
,
Oct 13 2017
The inconsistency happens when the value (e.g. "submitted") isn't one of the default options in the "Filter" textbox. "kicked_out" is one of the popped default options and the query is correct for "metric:action == kicked_out".
,
Oct 16 2017
Is this fixed? It no longer shows a constant value...
,
Oct 16 2017
The symptom fixed itself (something related to monarch query cache I guess), but in pcon I still see the "Not retained by any retention policy of role chrome-infra" message which is alarming and might have contributed to the problem.
,
Oct 16 2017
I don't know who else would take this. If anyone else *wants* to, feel free. :)
,
Oct 16 2017
It is retained, but under a slightly different set of target fields: http://shortn/_cfv95SPL3I Our retention policy is different for host_name ~= autogen:.* metrics vs. the non-autogen ones. Pcon may get confused trying to figure out the policy due to target fields being in the mix. So, I'd say it's WAI.
,
Oct 17 2017
|
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by akes...@chromium.org
, Oct 13 2017