Today I was updating VMware Site Recovery Manager (SRM) for a customer after their EMC VNX arrays were upgraded to a new firmware level.
It all went smooth and SRM seemed pretty happy with all green checkboxes, so it was time to run the Test Recovery Plan. But then the Test Recovery Plan failed, and returned the following error : “Error – Failed to recover datastore ‘DatastoreName’. An error occurred during host configuration”.
And another beautiful “An unknown error has occurred” message appeared in the Task & Events tab of the ESXi hosts where SRM tried to configure the NFS Export on.
According to the message the issue seems to be that the NFS Export couldn’t be mounted on the ESXi host, which is weird because a. the NFS exports could be mounted before and b. the hosts have active mounted NFS exports from the same storage array.
After some troubleshooting and digging through the log files the only thing that seemed different than before was that now the IPv6 addresses were added to the “Access hosts” list of the fail-over test NFS Export that SRM created. This is because SRM queries the ESXi hosts for all (!!) VMkernel ports and add those IP addresses to the “Access hosts” of the fail-over test NFS Export, including the IPv6 addresses.
Fortunately for me this customer doesn’t use IPv6 internally so it was quite easy to test if disabling IPv6 would solve the issue. And it did! After disabling the IPv6 the NFS Exports were mounted and the VMware SRM Recovery Plans were finished successfully.
IPv6 can disabled on a v5.x or v6.x ESXi host by running the following command from the CLI :
# esxcli network ip set --ipv6-enabled=false
Or through the Web Client : Host -> Manage -> Networking -> Advanced -> Edit -> IPv6 Support -> Disabled
Or just from the good old vSphere Client : Host -> Configuration -> Networking -> Properties -> Uncheck “Enable IPv6 support on this host system”
And then reboot the host.
We have submitted a SR to EMC and I’ll update this article if we receive any feedback on this issue.