1274355 - investigate replacing bugzilla cron with direct calls to bz

Status: RESOLVED WORKSFORME

(Socorro :: General, task)

Description

[Lonnen :lonnen]

Bugzilla association was initially implemented as a hack on a freeform field using a cron and a list off all the changes in bugzilla since the last cron run. It's expensive, and the DB schema for housing bug associations is oddly complex, with foreign key relations and at least one totally unused table.

These days bugs have their own field and we could look up bugs for a signature using the bugzilla api. Let's estimate how much traffic that would throw at bz, if caching would help much, and what perf hit we would take for doing that.

Comment 1

[Lonnen :lonnen]

Peter replicated the same calls we make in the webapp using an AJAX query to bugzilla.m.o. By his estimate, we would send about 1,000 calls per day from report/index. Topcrasher would add a few more, but those are easy to cache. Add about a hundred for signature report.

We could implement this in our model layer and have it transparently retrieve and cache, dropping the PG stuff behind it without much modification to the rest of the app.

It creates worse perf on a cold cache but otherwise is negligible. This is not something we likely want to do now, but will be a viable way of eventually dropping bug associations from c-s.

Comment 2

[Peter Bengtsson [:peterbe]]

Adrian, I shared the same thoughts on this with lonnen. 
He agrees that the solution is to just "pause" on the effort. It's just good to know that it's quite easy to accomplish a direct connection to bugzilla without the need for cron jobs and after-the-fact lookups of bugzilla metadata.

(note-to-self: the branch with the code is: load-bugzilla-by-signature-directly-from-report-index)