From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Chris <chris(at)bitmead(dot)com> |
Cc: | Matthias Urlichs <smurf(at)noris(dot)net>, pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: Performance (was: The New Slashdot Setup (includes MySql server)) |
Date: | 2000-05-19 16:19:24 |
Message-ID: | 8297.958753164@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general pgsql-hackers |
Chris <chris(at)bitmead(dot)com> writes:
>> not helpful at all WRT this benchmark, because the user who caused the
>> problem ("test", in my case) isn't permitted to run VACUUM on the
>> pg_index table.
> Speaking of which, why can't any user who can change meta-data, also
> Vacuum meta-data ? It's not a threat to security is it?
No, but it is a potential route to a denial-of-service attack, because
VACUUM has to acquire an exclusive lock on the target table. An
unprivileged user can't vacuum pg_index for the same reason he can't
lock it: he could effectively shut down all other users of that
database, at least for a while (and VACUUMs issued in a tight loop
might manage to make things pretty unusable).
The design assumption here is that VACUUMs will be run periodically by a
cron job executing as user postgres; typically once a day at a low-load
time of day is a good plan.
There has been some talk of switching away from the no-overwrite storage
manager to a more conventional overwriting manager. That'd reduce or
eliminate the need for periodic VACUUMs. But currently, you can't
really run a Postgres installation without 'em.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Richard J. Kuhns | 2000-05-19 16:25:35 | [HACKERS] Re: Question about databases in alternate locations... |
Previous Message | Michael Ma | 2000-05-19 16:13:02 | Why is the JDBC driver re-arranged? |
From | Date | Subject | |
---|---|---|---|
Next Message | Richard J. Kuhns | 2000-05-19 16:25:35 | [HACKERS] Re: Question about databases in alternate locations... |
Previous Message | Culberson, Philip | 2000-05-19 16:08:35 | RE: Question about databases in alternate locations... |