Version -- Oracle Solaris 10 1/13 s10s_u11wos_24a SPARC
Running as a ldom on T5-2 server, runs rac node of 11.2 oracle db
Connected to flash SAN storage
Dear All,
We observe spikes of I/O waits only on one node of a rac cluster database .
The load on the db is spiked with I/O waits at these times although generated work load remains the same. (Active session spikes from normal 8-10 to 40-50 causing OLTP to fail as well)
Main observation from iostat -xn 3 and sar -d 2 10 are service times spikes on this particular node without any considerable I/O load. (SAR outputs attached)
iostat -xe does not show any errors, and as per sys admins, service time delay shown on node is not shown in the SAN switch or in the array.
Normal service times remain less than 4 for very large IOPS (40000+) in node 1 but in node 2 IOPS of ~10000 would show service time of ~15-20. But this is not consistent as some time node two also
show large IOPS in normal service times.
Thanks in advance for any hint on same.
