Comment 5 for bug 424671

Revision history for this message
Martin Pool (mbp) wrote :

jtv just asked me about something like this: we are seeing many oopses caused by attempts to read attachments off big bugs (eg bug 88746).

From a Python shell you can fairly easily reproduce this with

lp.bugs[88746].attachments[0]

which gives something like:

https://lp-oops.canonical.com/oops.py/?oopsid=OOPS-1660S153

This oops is pretty pathetic because about 50ms all the actually necessary queries are done, then it spends the rest of its timeout doing completely unnecessary and painful other queries.

I would speculate that something in lazr's toDataStructure is interacting badly with storm.