Closed
Bug 1453249
Opened 7 years ago
Closed 7 years ago
[esr60] Cannot GET /json-automationrelevance/$HASH: Internal Server Error
Categories
(Developer Services :: Mercurial: hg.mozilla.org, defect)
Developer Services
Mercurial: hg.mozilla.org
Tracking
(Not tracked)
RESOLVED
WORKSFORME
People
(Reporter: jlorenzo, Unassigned)
References
Details
STR:
* GET https://hg.mozilla.org/releases/mozilla-esr60/json-automationrelevance/2c000486eac466da6623e4d7f7f1fd4e318f60e8
Results: 500 Internal Server Error
This prevents Taskcluster's taskgraph to generate optimized graphs. I think it blocks bug 1452958.
I'm not if bug 1441340 comment 5 will solve this or not. If so, please feel free to mark it as a dupe.
Reporter | ||
Comment 1•7 years ago
|
||
Hi :gps! Just to confirm, does bug 1441340 comment 5 solve this bug? Should I mark it as a dupe?
Flags: needinfo?(gps)
Comment 2•7 years ago
|
||
json-automationrelevance uses the pushlog to determine what files were touched by the push which includes the revision it was passed, and the pushlog is empty. So either some other not-written comment in that bug fixes it if we clone the pushlog when we clone a repo, which I don't think we do, or, it'll be fixed by having an actual push to mozilla-esr60.
Comment 3•7 years ago
|
||
This was due to the hgweb replicas being out of sync for an extended period of time.
As explained at https://mozilla-version-control-tools.readthedocs.io/en/latest/hgmo/notifications.html#common-properties-of-notifications, there is a race condition on every push where each hgweb read-only mirror obtains the new data at a different moment in time. We don't guarantee that https://hg.mozilla.org/ exposes a consistent snapshot of a repository until the Pulse or SNS message is sent. Systems that react to events on hg.mo should be subscribed to one of these feeds instead of polling hg.mo (e.g. via the pushlog).
Even if consumers are polling for changes, the window between replication inconsistency is often very small - <5s. So by the time a Taskcluster task hits hg.mo, the data is completely replicated and the window is long over. But we had an event that was causing long inconsistency windows and this caused chaos for downstream consumers.
Status: NEW → RESOLVED
Closed: 7 years ago
Flags: needinfo?(gps)
Resolution: --- → WORKSFORME
You need to log in
before you can comment on or make changes to this bug.
Description
•