Hi,
I've grumbled occasionally, without providing a lot of information,
about sporadically abysmal response times, especially when background
cron jobs such as slocate, tripwire-check and makewhatis were running.
Symptoms shown by top were load averages of 3-6 while cpu
was 95-98% idle. Interactive response was bad - and, when ssh logins
kept timing out yesterday, I decided it was time
to look more closely at the problem.
It turned out to be incompatibilty between my hard drives.
My system drive was a Quantum 13GB drive running UDMA2 on a Dell
optiplex P2-266 (Intel PIIX-something chipset).
Tests with hdparm -t -T reported about 18-19 MB/s xfer rate.
I added a second drive for data - Seagate 30GB. The Dell cabling
is too short to connect the second HDD to the second ide channel
along with the CD, so I put it as slave (hdb) to the system disc (hda).
I tested it and got 20MB/s, so I figured it was OK in that configuration.
What I didn't do was recheck the system disk - until yesterday -
which was now giving 1 MB/s (booting single user I could get it
to 2 MB/s). All indications were that both drives were still configured
for UDMA-2. Disconnecting the slave drive got it back to normal.
The solution was to get a longer ide cable and put the 30GB data
drive as hdc and the cdrom as hdd.
Now slocate.cron still takes 9 sec user and 18 sec system time, but
the elapsed time is 5 minutes instead of 30!
And I don't notice it.
Cameron.
--
redhat-list mailing list
unsubscribe mailto:[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-list
