[OSM-dev] vacuum running amok - and me too ;)

[Walter Nordmann]

Paul Norman wrote
> Indexes aren't vacuumed, tables are. Are you vacuuming the table or
> rebuilding theindex?

autovacuum vacuumed planet_osm_ways, but when this comes to the gin-index it
starts growing. Saw this during same tests running yesterday.

<http://gis.19327.n5.nabble.com/file/n5836445/pidstat.png> 
scale left is in MB.

at 2:25 it starts using the gin-index planet_osm_ways_nodes.

> If you have autovacuum tuned aggressively enough there should be no
> need to manually vacuum the slim tables. The GIN indexes do tend to
> bloat, and this can be fixed by stopping updates, dropping the old
> index and building a new one.

just doing it - recreating index. fyi: it's about 133GB big and needs more
then 10 hours to recreate it.


> If you have autovacuum tuned aggressively enough there should be no
> need to manually vacuum the slim tables.

is that the reason? doing not enough vacuums so that this one is too big and
that is making problems?

but *why is vacuum ignoring maintenance_work_men* and bringing my system to
thashing? I think, THAT is  the real problem - and we from osm can't fix it
:(

6 more hours to wait for the index, database offline to users, diff-import
stopped for 35 hours. i'v had better days.

Regards
walter





-----
[url=http://osm.wno-edv-service.de/residentials] Missing Residentials Map 1.17[/url] [url=http://osm.wno-edv-service.de/plz] Postcode Map 2.0.2[/url]
--
View this message in context: http://gis.19327.n5.nabble.com/vacuum-running-amok-and-me-too-tp5836373p5836445.html
Sent from the Developer Discussion mailing list archive at Nabble.com.