We've seen some occasional issues with academics regrading quizzes and getting timeouts. These are typically on quizzes with > 100 participants and large number of questions.

I realise that there is possibly some (more) tuning that we can do to prevent the timeouts, but I wonder whether perhaps regrades should be done asynchronously on cron. I'm not sure of the potential issues that this could cause, so I'm just throwing it out there for your thoughts.

Moving to an asynchronous regrade would mean that users aren't sat hanging and waiting while their session is locked and they're forced to go and make a cup of tea so it would also give a perceived performance improvement - especially if we had some form of live polling to fetch the status (as proposed in the backup plan too).

Here's a PERF log output for one that failed last night. This quiz had 108 responding users with 47 questions:

[Thu Nov 08 17:16:53 2012] [error] [client 148.88.22.10] PERF: /mod/quiz/report.php?id=49677&mode=overview&attempts=enrolled_with&onlygraded=&onlyregraded=&slotmarks=1&sesskey=GXaxzvxnbD&regradeall=Regrade+all time: 148.166105s memory_total: 16582776B (15.8MB) memory_growth: 15768696B (15MB) memory_peak: 16994384B (16.2MB) includecount: 379 contextswithfilters: 1 filterscreated: 7 textsfiltered: 33369 stringsfiltered: 0 langcountgetstring: 1592 langcountmemcache: 1583 langcountdiskcache: 9 db reads/writes: 19807/43741 ticks: 14817 user: 3271 sys: 345 cuser: 0 csys: 0 serverload: 0.04 Session: 24.2KB , referer: https://modules.lancs.ac.uk/mod/quiz/report.php?id=49677&mode=overview

(We have now increased our timeout substantially so that they no longer see the timeouts)