Unity-NAS Failover between SP

Question

hello everyone,    I can imagine this has been asked a lot of times. I wanted to know what are the possible solutions to minimize the downtime generated when a NAS server fails over to other SP while the SP fails/reboots. It is cleary written in the Dell EMC Unity: NAS Capabilities: page 9 : HIGH AVAILABILITY On Dell EMC Unity systems, both SPs can be used simultaneously so no dedicated standby hardware is required. The peer SP acts as a hot standby, which actively services I/O but is also ready to take over additional resources if necessary. For example, if SPA fails, the NAS Servers along with their file systems fail over to SPB. There may be a short interruption to host access during this operation.      How big or small is this short interruption in real life scenario or NFS. ? Any solutions to avoid it completely ?   Many Thanks, Vivek

Rainer_EMC · Accepted Answer

typically its between 10s and 1min depending of the config

these are values that any NFS client will tolerate

In any case unless you are using soft mounts (which you shouldnt) an application using a file over NFS will NOT get any errors - it will simply wait for IO

The OS NFS client stack is responsible for retransmitting a NFS reqeust if it times out.
The most you will see if a "NFS server not responding" and "NFS server ok" in the clients log - which isnt an error messages but simply an information that its been retrying NFS requests

vivek_asml · Answer

Thanks alot for taking time to reply Ranier, really appriciate. can LACP help here, specially across SP, will LACP work on unity systems ?   Regards, Vivek

Rainer_EMC · Answer

no  LACP is only possible with interfaces that are on the same SP same as with any other computer system

vivek_asml · Answer

Thanks for the help..as always :) Does implementing LACP give me high availability ? Which means if the cable or switch/switchport is faulted, the other NAS server interface serves the shares and hosts doesnt see any 'hickup' ? Cheers, Vivek

Rainer_EMC · Answer

sure - LACP can cover port / cable / switch failure

there will still be the same NAS server and interface used

port means a physical Ethernet port
above the physical port is LACP or FSN - still MAC layer
interface is at the logical IP level
lots if interfaces usually share the same port or LACP/FSN

see the Unity High Availability white paper

video here https://www.youtube.com/watch?v=hCYBMtDCD-A

DELL_ChrisHolloway · Answer

Just to add to the above answer: NFSv2/v3 are stateless so will just wait for the export to come back online during a failover. NFSv4 is stateful in that both the client and the NAS server keep information about locks. Locks move between the SPs during a failover and are reclaimed by the client once they move over and the reclaim happens within the grace period (gpDuration defined as a svc_nas param). Default is 45 seconds last time I looked into it.

cadencep45 · Answer

NFS is stateless, so client connection should simply hang during failover

SMB is stateful, so its more of an issue, except SMB3.

SBM1 should not be used ( depreciated and security hole)

SMB2 unless your clients support SMB is probably being used. If so, I think the SMB protocol times out in 45 seconds. So in failover takes longer than that, then client reconnnection is required. Failover time ( on Data Movers ), depended on the number of open files, - not sure of Unity - EMC can fill in the blanks :)

SMB3 should support seamless failover. But you needs clients ( Windows 10, Windows server x - not sure of version ) which supports it.

There is an option on unity to refuse a client connection unless its SMB3 , its in smb shares, advanced, continuous availability.

#S

Unity

Was this post helpful?