1552973 - Do not send CrashIDNotFound messages to Sentry, or combine into one issue

Status: RESOLVED FIXED

(Socorro :: Webapp, enhancement)

Description

[John Whitlock [:jwhitlock]]

With the Sentry SDK, the logger.error call when crash data is not found is turned into a Sentry event:

https://github.com/mozilla-services/socorro/blob/b649e89fc6dc440adf78fccecc0a8f2327b97cf0/socorro/external/boto/crash_data.py#L68

In some cases, this seems to happen when a crash is viewed before processing is complete, and doesn't re-occur when the crash is processed.

Because the message is distinct, a new event is created each time it is encountered.

One way to handle the issue is to apply formatting in the logger:

self.logger.error('%(datatype)s not found: %(exception)s',
                  {'datatype': params.datatype, 'exception': cidnf})

This should use the format string to add new events to the same issue, avoiding creating many similar issues.

Another option is to change to .warning, so that it will not be auto-logged. The formatting change can be made at the same time.

A third option is to ignore messages from this logger.

Comment 1

[John Whitlock [:jwhitlock]]

Attachment: pr 4943: bug 1552973: Log missing boto files as warnings

Changes .error to .warning, and switches from a pre-formatted string to a format string.

Comment 2

[John Whitlock [:jwhitlock]]

jwhitlock merged PR #4943: "bug 1552973: Log missing boto files as warnings" in 8990d39.

Comment 3

[John Whitlock [:jwhitlock]]

There's an existing crash with no metadata in stage that was causing one of these issues. I reloaded that page, got a "crash not found" message in the browser, a line in the stage logs, and no event in Sentry. Ready for prod.

Comment 4

[John Whitlock [:jwhitlock]]

Deployed to production, these messages are no longer added as events in Sentry.