#1282 weird actions graph
Closed: Fixed 4 years ago by praiskup. Opened 4 years ago by praiskup.


Metadata Update from @praiskup:
- Issue assigned to thrnciar

4 years ago

Metadata Update from @praiskup:
- Issue tagged with: bug

4 years ago

Problem is that 146 actions do not have filled ended_on column even though they have failed result code. Since these actions were malformed by manual intervention. I propose just manually fill in the ended_on column in production so they will disappear from graph.

UPDATE action SET ended_on = extract(epoch from now()) WHERE ended_on IS null AND created_on >= 1581064423 AND created_on <= 1581064544;

So I did, thanks for the sql query!

coprdb=# UPDATE action SET ended_on = extract(epoch from now()) 
coprdb-# WHERE 
coprdb-#     ended_on IS null AND 
coprdb-#     created_on >= 1581064423 AND 
coprdb-#     created_on <= 1581064544;
UPDATE 146

Metadata Update from @praiskup:
- Issue close_status updated to: Fixed
- Issue status updated to: Closed (was: Open)

4 years ago

Since these actions were malformed by manual intervention

Do we know what manual intervention, so we can avoid it in the future?

Yes, because we have wrong processing of large batches of actions -- some project
set max-builds to 10, which effectively planned removal of many builds
overnight. Then I went and set state=failed on those actions (otherwise copr
would be in weird state for multiple hours, maybe days). I forgot to set
ended_on. The question is why "only" 156 actions got this inconsistency ...
But those actions are from single project, and match this theory.

We should have tool for canceling the actions ... or better, to fix the action queue
so there's priority (action queue has priority feature, copr just needs to set it
correctly).

Login to comment on this ticket.

Metadata
Attachments 1