Impact of SCSI delays (enabling Brocade zoning) on "live" Oracle

Impact of SCSI delays (enabling Brocade zoning) on "live" Oracle

Post by Vsevolod Afanassi » Sat, 20 Jul 2002 17:44:54



We plan to implement Brocade zoning on our SAN, and I am looking for
any info on impact of enabling Brocade zoning on "live" Oracle
database
with IO going on.

Brocade manual makes following statement
"Zones cab be configured dynamically. When onfiguring new zones:
* Data traffic on ports and devices is idle for a moment, then resumes
accordingto the new zining configuration. ..." I am mostly
concern about this delay in SCSI traffic - is it going to cause Oracle
databases any problems?

We don't have "test SAN", so any testing is rather difficult.
Nevertheless, I was able to conduct a test with enabling/disabling
zoning while
running "mkfile" or "cp" on SAN-based filesystem (no Oracle).
While the box (Sun2.8) did not crash and the filesystem remained
available,
I got messages in /var/adm/messages


(sd132)
SCSI transport failed: reason 'tran_err': retrying command

What impact such SCSI delay would have on Oracle (various versions
from 8.0.5 to 9.0.1)? Probably, ALTER SYSTEM SUSPEND will be helpful.