| SAN FAILOVER TAKES 110 SECONDS, WHAT IS THE IMPACT FOR AN INSTANCE? [message #203575] |
Wed, 15 November 2006 10:13 |
suvv
Messages: 17
Registered: October 2006
|
Junior Member |
|
|
We are performing a test with a new SAN (EMC CX3-80) on Solaris 8 using Oracle
9.2 and 10.2
We are using the EMC software called PowerPath 4.5.0 which performs load
balancing and failover. We have two HBA fibre channel cards SUN X6767A
(SG-XPCI1FC-QF2) connected to the SAN using HBA firmware 1.14.15 and the QLC
driver 11.8.0 which is part on the Sun bundle called SAN 4.4.10.
When we test the failover by disabling a port on the SAN the failover takes 110
seconds. There is no traffic on the remaining path during this failover. After
110 seconds we see traffic on the path. We also changed the Solaris parameter
fp_offline_ticker from 90 (the default) to 10 which reduced the interruption
from 110 seconds to 30 seconds. This was done after liasing with Sun technical
support.
The first tests were done using touch to insert the date and time in a files
during the failover. This test worked without any issues. The tests performed
using Oracle will be described below.
Oracle tests using Oracle 9.2.0.4, 10.2.0.1
1st test
----------
initiate failover
Shutdown immediate
no entries in alert.log for 110 seconds, no response from SQL> prompt during
the failover - for us this is not a good sign
2nd test
----------
initiate failover
Startup
no entries in alert.log for 110 seconds, no response from SQL> prompt during
the failover - for us this is not a good sign
3rd test
----------
initiate failover
alter system switch logfile;
no entries in alert.log for 110 seconds, no response from SQL> prompt during
the failover - for us this is not a good sign
I am Unable to deploy the SAN, unsure of impact for the use of high volume instances.
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
