extreme performance regression on boxset over the last 10 days



Camino Graveyard
9 years ago
9 years ago


(Reporter: Smokey Ardisson (offline for a while; not following bugs - do not email), Assigned: Samuel Sidler (old account; do not CC))


({perf, regression})


While boxset was offline for Sam's move, or as a consequence of Sam's move, there was a massive, across-the-board perf regression:

    Ts: ~1630ms -> ~1830ms (~12% regression)
    Tp: ~ 540ms -> ~ 950ms (~76% regression)
Tdhtml: ~ 850ms -> ~1230ms (~45% regression)

I don't know if there's anything we can do, if boxset is just broken, if there was a nasty checkin during the period, if some stray process is running, or what, but filing because it was so massive and so we have a record.


Of stuff that gets built, there's the Gecko milestone bump (bug 487028), the Sparkle update (bug 445629 comment 13), and a security fix involving XSLT somehow (bug 483444).  None of them seem particularly likely, though you could perhaps attempt to make an argument for a very small upward shift in the "normal" Ts "range" on cb-x1 around the time of the Sparkle update.
Whatever are you talking about? Perf looks great!

(Today, Mountain View experienced a power outage. After getting home and restarting boxset, tests seemed to return to their normal levels.)
Even looks like a preliminary Tp win :)

Finders crossed, we can close this sometime tomorrow if we keep steady.
It's been 24 hours now and we're still stable, so FIXED.  

(Assigning to Sam since I can't find a contact from his power company to credit for the 'fix'.)
Assignee: nobody → samuel.sidler
Last Resolved: 9 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.