Re: Maximum Performance Follow-up Question - Mailing list pgsql-admin

From Peter T. Brown
Subject Re: Maximum Performance Follow-up Question
Date
Msg-id 002101c1a5d8$c09b5990$7d00000a@PETER
Whole thread Raw
In response to Re: Maximum Performance Follow-up Question  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-admin
Actually, a great many of the queries we run are Very dependant on
inserts -- we do inserts into a big 'pointers' table and use that as a basis
for other queries. Its also the case that for our web tracking application
nearly every select is paried with an insert (lookup a visitor, add a row
recording their hit to the website). So I think that gaining efficiency on
inserts would really help...

Is there any BIG risk in turning fsync off? I mean, if I miss the last 30
minutes of tracking data from our website because of a system crash, thats
no big deal to me. If the entire database gets corrupted and must be
scrubbed, that's a big deal.

And isn't there some way to use fsync but just use it less frequently, so
that postgres writes a bunch of changes to disk when it less busy or
something?

Thanks

-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: Friday, January 25, 2002 7:24 AM
To: Radu-Adrian Popescu
Cc: Peter T. Brown; Postgres Admin List
Subject: Re: [ADMIN] Maximum Performance Follow-up Question


Radu-Adrian Popescu <radu.popescu@aldratech.com> writes:
> I belive you should set
>     fsync=false
> in case you mainly select and do inserts rather rare.

No, that's a really horrid reason to turn off fsync.  A read-only
transaction never syncs and thus has no fsync penalty.  If update
performance isn't a serious problem for you, you may as well keep
fsync on and not have to worry about data loss in the case of a
system crash.

            regards, tom lane


pgsql-admin by date:

Previous
From: Tom Lane
Date:
Subject: Re: Maximum Performance Follow-up Question
Next
From: Antony Stace
Date:
Subject: Logging in postgres