Thread: Best practice

Best practice

From
geustace@godzone.net.nz
Date:
Now that 7.1 has arrived, I would appreciate some feed back on the
following comments.

We use Ingres where I work and when setting up a server we always
try to ensure that the log and data files are on different spindles.

From both experience and advice from CAI, one can always recover a
database after a hardware failure or system crash, if any two of the
following are in good condition

Backup, Logfile, Datafiles.

I know that PostgreSQL isn't Ingres but there must be some useful
experiences from both experts and users alike. Is there any recommended
'placement' for the respective postgresql files.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Glen and Rosanne Eustace,
GodZone Internet Services, a division of AGRE Enterprises Ltd.,
P.O. Box 8020, Palmerston North, New Zealand 5301
Ph/Fax: +64 6 357 8168, Mob: +64 21 424 015



Re: Best practice

From
Tom Lane
Date:
geustace@godzone.net.nz writes:
> We use Ingres where I work and when setting up a server we always
> try to ensure that the log and data files are on different spindles.

This would be good practice in PG simply for performance reasons.
Ideally the WAL log should be on a disk that has nothing else to do,
so that you never have to seek somewhere else than the current WAL
segment.

I'm not sure that we can yet guarantee very much about recovery from
disk hardware failures.

            regards, tom lane

RE: Best practice

From
"Willis, Ian (Ento, Canberra)"
Date:
What would the the best choice for the WAL mirroring, raid3 or 5. How big
does the WAL grow? mirroring is fine as long as the size isn't too big.



--
Ian Willis

-----Original Message-----
From: Tom Lane [mailto:tgl@sss.pgh.pa.us]
Sent: Friday, 20 April 2001 9:22 AM
To: geustace@godzone.net.nz
Cc: pgsql-general@postgresql.org
Subject: Re: [GENERAL] Best practice


geustace@godzone.net.nz writes:
> We use Ingres where I work and when setting up a server we always
> try to ensure that the log and data files are on different spindles.

This would be good practice in PG simply for performance reasons.
Ideally the WAL log should be on a disk that has nothing else to do,
so that you never have to seek somewhere else than the current WAL
segment.

I'm not sure that we can yet guarantee very much about recovery from
disk hardware failures.

            regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/users-lounge/docs/faq.html

RE: Best practice

From
"Mikheev, Vadim"
Date:
> What would the the best choice for the WAL mirroring, raid3
> or 5. How big does the WAL grow? mirroring is fine as long
> as the size isn't too big.

1. For each transaction that made some updates to DB
   all log records must be available.
2. Server removes "old" (see 1.) 16Mb log files only at
   checkpoint time (each ~300 sec by default).

So, in the system with low update rate/short update TX
there will be only 1 - 2 16Mb log files at any time.

Vadim