Hi,
It looks like you are hitting bug
CSCti17353 .
Bug Details
"IBM Servers battery learn cycle negatively impacts CCM performance"
Symptom:
CCM Tracing stops for a period of time (over 4 seconds observed) leading
dropped or reordered calls.
IO wait utilization of the CPU increases unexpectedly.
System Event Logs (messages) contains the following logs shortly after the
unexpected IO event.
Aug 2 15:48:08 lg-sub-1 daemon 5 Director Agent: LSIESG_AlertIndication
500605B00177F4C0
Battery relearn started Sev: 2.
Aug 2 15:48:13 lg-sub-1 daemon 4 Director Agent: LSIESG_AlertIndication
500605B00177F4C0 BBU
disabled; changing WB virtual disks to WT Sev: 3.
Conditions:
On MCS-78X5-I3 platforms with LSI Raid controllers the system will run a
Battery Backup Unit
learning cycle every 60 days. The system can experience sudden IO
performance drop at the beginning
of this cycle and this could impact Callmanager's call processing
performance. During this cycle the
Raid Controller's Caching policy is changed from WriteBack to Write Through.
Workaround:
None to avoid the BBU learning cycle.
After this ddts is applied...
The event reported above is an informational (sev 4) event: daemon 4
Director Agent:
LSIESG_AlertIndication 500605B0027DA6E0 BBU disabled; changing WB virtual
disks
The event will occur once a month (at 4:42am) during the battery recharging
cycle. That cycle is a
scheduled event to recharge the raid battery. During that time, raid cache
write-back is disabled as
part of normal course to ensure proper IO. This is not an issue.
