HP StorageWorks SAN Virtualization Services Platform Best Practices Guide (5697-0935, May 2011)

Operating system-specific monitoring
VMware recommends that very large I/O sizes be used for data transfers. These recommendations
are valid when the storage is local to the servers; it does not apply to array controllers on the SAN,
specifically SVSP. Check ESX servers to ensure that all paths to the DPMs are visible. If the zoning
is correct, you may need to perform a rescan several times to ensure that the path is set up properly.
Consult the monitoring sections of your OS documentation on how to monitor common Fibre
Channel errors. Usually the HBA provides tools for observing the Fibre Channel interface.
Monitoring setup virtual disk performance
There are two event log messages related to the setup virtual disks that will indicate if there are
issues. Response time is averaged on the setup virtual disks every 10 minutes, and one informational
message is published each day. Monitor these messages daily to determine whether the setup
virtual disk response time needs to be improved. The event log messages that the system reports
for synchronous mirror groups and tasks are also generated for the synchronous mirror group of
the setup virtual disk.
The messages indicative of a slow setup virtual disk are:
• "Sync.Mirror [Setup_volume_job_name]: average write response time(X msec) >= 50 msec"
X is the average write time of all the 10-minute intervals when the average exceeds 50 msec.
• "Sync.Mirror [Setup_volume_job_name]: X writes took more than 5000 msec" - logged every
10 minutes if there were 1 or more I/Os to the setup virtual disk during the past 10 minutes
which took more than 5 seconds.
Capacity monitoring
There are global and individual mechanisms to ensure that operations are not curtailed because
a storage pool is running out of free capacity.
Global mechanisms
• Warning (10% free)
Notification to event log every 5 minutes◦
◦ Delete oldest asynchronous mirror PiTs
◦ Keep most recent asynchronous mirror PiT
• Guard (5% free)
Suspend asynchronous mirror tasks◦
◦ Delete most recent PiT
◦ Delete migration tasks
◦ Delete oldest PiT
• Emergency (2 % free)
◦ Delete all PiTs (starting from the oldest)
Individual mechanisms
Percentage on an individual virtual disk
When free capacity drops below the threshold:
• Notification occurs every five minutes
• PiT creation is stopped
Monitoring for performance 59