1432995 - Sync telemetry should report more detail about http errors

Status: RESOLVED FIXED

(Firefox :: Sync, enhancement, P1)

Description

[Mark Hammond [:markh] [:mhammond]]

Looking at sync telemetry, we see a large number of errors reported as ("othererror", "error.engine.reason.something"). In almost all cases, this tends to mean an unexpected http status code (other than 5xx errors which are reported). This makes it difficult to determine what might be going wrong.

For example, on some days, over 5000 unique devices see "record_upload_fail" doing the history engine, and this doesn't seem to correlate with stats from the server  - having the status code is likely to give us more insights into what the problem actually is. These errors seems transient, but still aren't ideal.

We should even do this for "error.engine.reason.batch_interrupted" - having this reported as ("httperror", 412) gives us the same information.

Note that doing this would mean we lose information about whether the problem was during upload or download. That seems a reasonable tradeoff for a first cut - we could consider adding the http method used in another bug if we found it would be useful.

Comment 2

[Thom Chiovoloni [:tcsc] (ex-moco)]

Attachment: Bug 1432995 - Check for HTTP error codes before error strings in the sync ping.

Review commit: https://reviewboard.mozilla.org/r/216326/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/216326/

Comment 3

[Mark Hammond [:markh] [:mhammond]]

Comment on attachment 8946381 [details]
Bug 1432995 - Check for HTTP error codes before error strings in the sync ping.

https://reviewboard.mozilla.org/r/216326/#review222270

Awesome, thanks!

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/7bc2c9f1383e
Check for HTTP error codes before error strings in the sync ping. r=markh

Comment 5

[Cosmin Sabou [:CosminS]]

https://hg.mozilla.org/mozilla-central/rev/7bc2c9f1383e