We have four hardware identical Cisco UCS servers that installed as Exchange mailbox servers mailbox role and joined DAG replication partners. Two local servers are working fine. However, two remote servers keeping running out of memory and we had to reboot these two servers at regular base (once a month) until we used Poolmon to identify the cause of the memory leaking, a known Cisco NIC teaming bug. Poolmon was able to identify with tag Team caused memory leaking that consumed more than 5 GB nonpaged memory before the server running out of available memory less than 20MB of 96 GB physical RAM and Hard Faults per sec pager constantly reached 100.
Once poolmon identify tag Team, then we used the following string to find the driver that caused the memory leaking:
findstr /m /l team *.sys either of system32 or \system32\drivers directory.
for more info, please read the following links:
No comments:
Post a Comment