SCSI-array questions

Gabor Kondorosi gabor.kondorosi at mizuho-sc.com
Sun Mar 30 21:21:01 CST 2003


Dear All,
 I have PV220S configured in cluster mode shared by two PE2650s with a
PERC3/DC in each.
There are two RAID-5 containers (6 disks each) set in the array with a
global HS.
OS is RH 7.3 (2.4.18-27.7xsmp kernel), the PV 220S firmware is E11.

First of all, I found that the I/O performance is simply terrible.
Bonnie++ gives 4MB/s write I/O (!!!) and 20MB/s read I/O. During heavy
write operations the machine practically hangs for 5-10 seconds. I do
not see SCSI timeouts or anything relevant in the syslog. Having a
RAID0+1 configuration does not seem to improve the performance at all.
In comparison: PERC3/Di generally gives 35 / 70 MB/s (Write /Read; U160,
RAID-5) while a PERC4/Di in a PE2600 gives 70 / 110 (U320, RAID-5).

Second: MegaMon seems to die (an issue mentioned earlier by someone)
after exactly 8 hours and has to be restarted every time. I could not
find any default value in the MegaCtrl help that would imply this.
[MegaCtrl is started as 'MegaCtrl -start' by the default rc-startup
script from the RPM-package]. The version used is: DELL Version : 3.6
Jun 13, 2002.
Is this a normal behavior?

Third: This was the second occurrence that about a week ago the
high-temperature condition LED went off on the front panel of the PV220S
and it has been on continuously since then although I'm sure the
temperature is back to normal again. Also, both fans are working like
crazy. I'm not aware of any Linux tool that would allow me to check the
readings of the temperature sensors inside the array and re-set this
condition. Is there any other way to solve this than a power-cycle of
the disk array (it's a production environment)? Btw, this problem seems
to appear after upgrading the firmware (didn't have this problem with
E10).

I would appreciate any feedback from you on these.
Thanks in advance!
--Gabor




More information about the Linux-PowerEdge mailing list