Alvaro Herrera wrote:
andy wrote:
with autovacuum enabled with default settings, cramd.sql is 154M:
[EMAIL PROTECTED]:/pub/back$ time pg_restore -Fc -C -d postgres cramd.sql
real 3m43.687s
[...]
Now I dropdb and disable autovacuum, restart pg:
[EMAIL PROTECTED]:/pub/back$ time ( pg_restore -Fc -C -d postgres cramd.sql;
vacuumdb -z cramd )
real 3m47.229s
user 0m9.933s
sys 0m0.744s
Sweet, about the same amount of time.
Thanks. I find it strange that it takes 3 minutes to restore a 150 MB
database ... do you have many indexes?
Even though the restore times are very similar, I find it a bit
disturbing that the "CREATE INDEX" is shown to be waiting. Was it just
bad luck that the ps output shows it that way, or does it really wait
for long?
There are about 800 tables, each has one index. Most tables (75%) are
very small, the rest have, maybe 50K rows.
I had to run the ps several times to catch it waiting. It didnt seem to
wait too long.
It was run on my laptop, which may not have the best io times in the
world (and it only has 512 Meg ram).
-Andy
---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster