Re: 8.3 / 8.2.6 restore comparison - Mailing list pgsql-hackers

On Thu, 07 Feb 2008 19:20:26 -0500
Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Looks to me like you're disk-bound (and your kernel is pretty bad at
> distinguishing "idle" from "disk wait" states).  Plain COPY into an
> index-less table ought to be writing fairly linearly, so I'm surprised
> you aren't able to get more out of the disk.  Have you got WAL on
> a separate spindle from the database?

I didn't, but do now and am testing. The way this is currently designed
is:

/data1 (8 disks RAID 10)
/data2 (8 disks RAID 10)
/pg_xlogs (2 disks RAID 1)

/data1 is what is primarily written against for the first couple of
hours and then it will switch to data2 because of table spaces. However
either way, we should expect (I would think) to see at least 100 megs a
second on an 8 Disk RAID 10. It is SCSI.

vmstat -adSK 5
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----r  b   swpd   free  inact active   si
so   bi    bo   in    cs us sy id wa1  1    160 6534048 23711184 1434532    0    0  1571   387    0     0 10  7 71 121
1   160 6206560 23898644 1573680    0    0 18168 23851 2684  1647 11  3 84  21  2    160 5887072 24080840 1711060    0
 0 17655 58395 3322  1762 11  3 74 112  2    160 5546160 24275268 1856108    0    0 18681 61834 3302  1838 11  3 74 111
2    160 5363432 24383836 1933028    0    0 13043 52353 3247  1578 12  2 74 111  1    160 5295336 24423764 1963808    0
  0  4997 39778 2865  1310 12  1 75 121  0    160 5284184 24426444 1971608    0    0   385  6967 2087   955 13  1 84
31 0    160 5274608 24429180 1978236    0    0   384  1009 2030   958 13  0 87  01  0    160 5265408 24431976 1984688
0    0   410   946 1780   656 12  0 87  01  1    160 5224912 24453296 2003700    0    0  1998 23168 2419  1163 12  1 79
81  2    160 5005048 24605300 2069236    0    0 18270 41342 2942  1721 12  2 76 101  2    160 4816016 24748120 2117208
 0    0 19014 37165 3014  1702 12  2 72 151  1    160 4618064 24894476 2167412    0    0 19091 15226 2517  1626 12  2
83 31  1    160 4423376 25039072 2218324    0    0 18603 10398 2585  1644 12  2 84  31  1    160 4223568 25187592
2268816   0    0 19475 10326 2511  1666 12  2 84  21  1    160 4023056 25337080 2319844    0    0 19450 10601 2462
154312  2 84  21  1    160 3821328 25487664 2370328    0    0 19834 10657 2454  1540 12  2 84  21  1    160 3617616
256408322420820    0    0 20398 11202 2536  1678 12  2 84  21  2    160 3410040 25795488 2473800    0    0 20090 29063
2756 1680 12  2 76 101  1    160 3189064 25957688 2531996    0    0 20321 13745 2523  1595 12  2 83  4 

Sincerely,

Joshua D. Drake


--
The PostgreSQL Company since 1997: http://www.commandprompt.com/
PostgreSQL Community Conference: http://www.postgresqlconference.org/
Donate to the PostgreSQL Project: http://www.postgresql.org/about/donate
PostgreSQL SPI Liaison | SPI Director |  PostgreSQL political pundit