HPE SimpliVity
cancel
Showing results for 
Search instead for 
Did you mean: 

Re: Failure scenario with arbiter

 
SOLVED
Go to solution
Highlighted
Frequent Advisor

Failure scenario with arbiter

Hi, 

knowing that article from vmware : https://kb.vmware.com/s/article/51462

But what happen in the following scenario (sur that BP is to have arbiter on a third site) :

Server 1 in site A

Server 2 in site B

Arbiter in site B

If site B is isolated from A, VM continue to run on server 2 in B, and VM that were on  server 1 in A are restared on B site, correct ?

If site B is lost (server 2 and arbiter down) what happen to site A ? I suppose that A will be in Data Unavailability because not able to reach Server 1 and Arbiter in B, correct ?

Thanks for your help,

Regards.

Jean-Philippe

 

4 REPLIES 4
Highlighted
HPE Pro
Solution

Re: Failure scenario with arbiter

Correct. The nodes in Site A will go "Read Only", because they will believe they have been isolated from the network (unable to contact other nodes, unable to contact Arbiter).

That's why it is important to have the Arbiter in a third site (which coudl eb the cloud if your latency is low)

Note that the Arbiter can be moved , so if necessary, you can deploy it in Site A or B in order to get everythig running - just don't forget to move it to a third site to make the cluster fully fault tolerant <g>.


I work for HPEAccept or Kudo
Highlighted
Frequent Advisor

Re: Failure scenario with arbiter

Thank you John.

Highlighted
Frequent Advisor

Re: Failure scenario with arbiter

By the way, what about returning to a nominal situation ? Does it mean that to get back the ESXi working the failed nodes and arbiter must be back online to allow the "surviving" node working Read / Write or is there a  solution to allow the last node to be able to work ? 

If Node and arbiter are down for 3 days the last runing node in Read Only can be set to RW by the support ?

Hope to be clear.

Thanks for your help.

Highlighted
HPE Pro

Re: Failure scenario with arbiter

Case 1 : If DRS is manual ; VM's needs to be manually migrated to the compute of Site A while storage continues to stay in the existing NFS.

 

Case 2 : If DRS is partial /fully automated ; VM's will continue to stay on the compute of site B or migrate to the compute of site A based on the DRS rules.

 

Arbiter : Arbiter has no role play in VM's or OVC failback.


I work for HPE

Accept or Kudo