Please share your experience about the impact on services when failover/failback [ Edited ]

[#1]

We have heavily loaded NFS v3 mounts, vmware, and CIFS shares on cDOT cluster with 2 nodes. If we perform failover and then failback practice, services should not be interrupted. Can anybody please share your expriences on what symptom, any possible slowness, or any kind of impact? are they truely no interruption on Database, Website, etc conections?

Appreciate your sharing.

[#2]

Unless you are running Continous Availability SMB 3.0 shares, all your cifs traffic will be disruptive

[#3]

NFS and vmware should be transparent if configured suitably. There will pause in accessing data from NetApp so all applications need to be able to survive it. You can look at documentation for various host utilities which timeouts they set by default (they should represent worst case).

CIFS sessions will be interrupted and applications will need to reconnect.

[#4]

we are running Redhat and Windows on ESXi host, please advise what Host Utilitiy we should install, and it'd be great if you can forward link to the software. Thank you, and hear you soon.

[#5]

Also, what about iSCSI? Same as CIFS, do we expect the interruption as well since both are stateful interruption?

If iSCSI is interruptinve as well, then we may have issues since ESXi hosts are using iSCSI lun's for booting.

[#6]

With iSCSI host should transparently reconnect, just like with NFS, if it is configured following NetApp recommendations.

[#7]

Hi aborzenkov ,

I guess, what you are referring to is that all iSCSI Luns should have the timeout set to 2 mins, and that is NA recommendation. If we don't have this parameter set, and then to change it, does that require to reboot the host? What is the command or the method to verify the parameter?

Another question, assuming the pause would be longer than the timeout, since iSCSI LUN in this case is bootable drive, will we then have any issues if we don't boot during failover/failback?