Re: NFS mounted DBs and Vacuum - Mailing list pgsql-general

From Steve Wolfe
Subject Re: NFS mounted DBs and Vacuum
Date
Msg-id 009e01c07031$89b19940$50824e40@iboats.com
Whole thread Raw
In response to NFS mounted DBs and Vacuum  (Webb Sprague <wsprague@o1.com>)
List pgsql-general
> I have a server for which PG_DATA = /home/maxtor, an NFS mounted
> disk.  Normally it works great, but when I try to vacuum, it takes
> FOREVER.  Should I not even try to use remote storage like this?  Has
> anybody else run into a similar problem?

  NFS is slow, and very CPU-intensive.  On a 100-megabit switched link with
multiple, fast processors on the NFS machine, you get up to a maximum of
around 1/4 of the throughput that you would having a single, fast drive in
the database machine.  A handicap like that is bad for a database. : )

   I imagine that you are mounting the data directory from a file server for
backup purposes and disk redundancy protection.  We spent a little more
money, and put another RAID array in our database server for disk
redundancy, and each night a pg_dumpall runs, with the output gzipped and
THEN copied via NFS to the file server, to be included on the nightly DAT
backup.

steve



pgsql-general by date:

Previous
From: Webb Sprague
Date:
Subject: NFS mounted DBs and Vacuum
Next
From: "Robert B. Easter"
Date:
Subject: Re: LD_LIBRARY_PATH