Sean Crookston

Those that have used any form of the Celerra VSA know that at times it can be difficult to troubleshoot issues as result of it being a simulator and not the real thing. Best practice when issues occur is to revert the build back to the original configuration and try again, but recently I began having issues when setting up VMware Site Recovery Manager even after doing so.While I wish to say the root cause of the problem was some elaborate issue it was not. Simply put I failed to properly size the source filesystem containing the iSCSI lun that was being replicated.

While replication worked fine as result of the initial deployment not containing any real data or data that would have changed, running a test failover in SRM resulted in the error below:

Non-fatal error information reported during execution of SRA:testFailover Output: filebase: command, dir: C:/Program Files/VMware/VMware Site Recovery Manager/scripts/SAN/celerra/, fileExt: .pl Error:”

Not being an expert on the topic I found it difficult to troubleshoot this issue and what was even more frustrating was that performing an actual failover DID work. Upon discussing with a co-worker the issue it led me down the path of sizing, where I found this document from EMC on sizing specifically for iSCSI luns on the Celerra. After sizing out the source lun correctly the test ran correctly. While I haven’t been able to yet figure out what the minimum requirement of SRM is to successfully run a test I do know that giving the source lun the appropriate amount of storage using the calculation below resolved the issue.

This document goes into great detail sizing out for Celerra replication using iSCSI luns and even takes into consideration whether the source filesystem and iSCSI luns are virtually provisioned or not, which can have quite a different effect on the resultant size of your sizing requirements.

The equation to use for sizing is below:

FSmin = SizeofLun + SizeforSnapshots + SizeforTWS (Temporary Writable Snapshot)

Simply put the size out lun is going to be the size of the iSCSI lun plus the size for snapshots(cumulative) plus the size of TWS, which represents the changed data . These number can vary greatly depending on utilization, but two important things to recognize are that the SizeforSnapshots and the SizeforTWS can be quite large. In fact you should plan to allocate at least the size of the iSCSI lun for the initial snapshot alone on the source files sytem, plus you’ll need to take into consideration other snapshots and the temporary TWS space.

For those without an EMC Powerlink account I also found this blog that goes into much more detail on sizing iSCSI for the EMC Celerra.