It's part of the conventional wisdom of performance tuning to keep
pg_xlog on a separate physical disk, and this seems to be common sense
to me. Am I correct in deducing that the performance gain comes from
preventing a file being written to sequentially from competing for I/O
resources with the data presumably experiencing somewhat random access
in terms of reads/writes? If so, would it be useful to add a few
explanatory sentences to the docs?
http://www.postgresql.org/docs/7.4/static/wal-internals.html
"It is of advantage if the log is located on another disk than the main
database files. This may be achieved by moving the directory pg_xlog
to another location (while the server is shut down, of course) and
creating a symbolic link from the original location in the main data
directory to the new location."
Anyway, I'm wondering what experiences people have had in setting this
up in production. If I have a server with 5 drive bays, one thought I
had was to have a RAID 1+0 for the primary partition and then to put
WAL on the fifth disk, but I assume that failure of that drive if there
were no redundancy would be a Bad Thing. Another option would be to do
RAID 5 on three drives and WAL mirrored on the remaining 2 drives. But
RAID 5 is slower than RAID 1+0 for data I/O in that scenario according
to what I've read.
Would a more ideal setup be a six-drive server with a RAID 1+0 on four
drives with 2 drives dedicated to WAL that mirrored one another? Is
that overkill? Anyone have recommendations for an appropriate setup for
a 5-bay server?
Thanks!
-tfo
--
Thomas F. O'Connell
Co-Founder, Information Architect
Sitening, LLC
http://www.sitening.com/
110 30th Avenue North, Suite 6
Nashville, TN 37203-6320
615-260-0005
