Infrastructure

Florian Effenberger wrote:

Do you have an exact date and time, so we can see in the server logs?

Florian, for one, it's still giving me an "Internal server error" right now, that's Tuesday, January 17 at 5.01 pm GMT, when I enter something, anything, in the "Search or Ask your Question" box at the top, and then hit enter to search. See attached.

Hmm, I noticed there were dozens of thousands of dead rows in the "django_session" and (more surprising) "askbot_thread" tables. I did a manual VACUUM FULL, but only 2.5h later we back at

INFO:  "django_session": scanned 30000 of 61128 pages, containing 1292950 live rows and 5141 dead rows; 30000 rows in sample, 2640563 estimated total rows
    INFO:  "askbot_thread": scanned 3187 of 3187 pages, containing 25612 live rows and 3247 dead rows; 25612 rows in sample, 25612 estimated total rows

Evgeny, the number of rows in "django_session" is surprisingly high isn't it? If that's expected I guess storing them in shared memory instead would improve performance? Also, I'm a bit puzzled at the ever increasing dead rows in "askbot_thread". Is that normal behavior? Perhaps we just need a more aggressive autovacuum policy on these tables?

Is this ticket dealt with?

I didn't hear loud complains in a while, but I think the PostgreSQL database needs a more aggressive VACUUM policy on some tables to avoid clogging up the database. I'm still monitoring the progression of dead rows to see how to fine-tune that policy.

Closing, as this seems to be solved now