User Tools

Site Tools


Cannot create a Protection Group using SRM and EMC VNXe and NFS

So that was a nice case I worked on with a customer, and I documented the issue in a KB here

The issue was, even though we see the replicated devices/shares, you cannot see any datastore groups during PG creation. The NFS shares were fully accessible from ESX for the root user, so we did not have a R/W or permission issue. What the hell could it be?

Off to the logs, as we say here, logs are truth.

SRM logs showed something rather interesting:

--> **** Response ****
--> <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
--> <Response xmlns="">
-->     <ReplicatedDevices>
-->         <StoragePorts>
-->             <StoragePort type="NFS" id="XX.XX.XX.XX"/>
-->         </StoragePorts>
-->         <TargetDevices>
-->             <TargetDevice state="read-only" key="100_SOMETHING1_0000_108_SOMETHING2_0000">
-->                 <Name>Performance01</Name>
-->             </TargetDevice>
-->             <TargetDevice state="read-only" key="116_SOMETHING3_0000_104_SOMETHING4_0000">
-->                 <Name>SRM-Replication</Name>
-->             </TargetDevice>
-->         </TargetDevices>
-->     </ReplicatedDevices>
--> </Response>

Did you notice it? No?

Take a second look at the TargetDevice state that the SRA is returing. That's right, the state is read-only. WTF?

We head back to Unisphere, and sure enough, the NFS ACL default is set to R/W+ROOT, and all the Hosts assigned to it are using “Default”, meaning they inherit the R/W+ROOT setting, which is why ESX is working without any issues.

So how come the SRA is returning read-only?

To test, we set the hosts ACL explicity to R/W+ROOT and tried again from SRM, and poof, it worked.

Is this a bug in the SRA? Is the array return different values depending on who is asking?

My guess is that the SRA sees “Default” and automatically sets read-only. The correct behaviour would be to go up the chain and find out what “Default” is. It is makign an assumption, a bad one.

So I *fixed* it, but this is really a workaround. I am hoping someone at EMC will fix this. We do not see that many VNXe and NFS, but they will get popular at one point.

So there it is, use the KB link this it is the official one, and give me some feedback :)

vmware/srm_vnxe_nfs_pg.txt · Last modified: 2013/05/10 20:54 by naccad

Donate Powered by PHP Valid HTML5 Valid CSS Run on Debian Driven by DokuWiki