3PAR StoreServ Storage
cancel
Showing results for 
Search instead for 
Did you mean: 

3PAR Peer Persistence with VMware Stretch Cluster

jose da silva
Frequent Advisor

3PAR Peer Persistence with VMware Stretch Cluster

Hi 

Need advice on the 3PAR Peer Persistence the best practices to shutdown one site of an Active/Active "VMware 6.0" Stretched Cluster. Taking into concideration vMotion of the vms from one site to the other and then shutdown one site for maintenance.

RegardsJose

3 REPLIES
ArunVC
Visitor

Re: 3PAR Peer Persistence with VMware Stretch Cluster

hello,

To my understanding when the site itself goes down, this will trigger an autofailover once the primary site is down and you VM's will move to DR site. 

Before teh DR you will need to see if all the ESX are logged into storage and on the ESX end you will need to check if all the luns are able to see all the active and standyby paths

StorePete
Occasional Visitor

Re: 3PAR Peer Persistence with VMware Stretch Cluster

I assume you do have Peer Persistence set up with Quorum Witness for transparent failover support. Here are the basic steps:

1. Stop the Quorum Witness

Find the procedure to stop and start 3PAR Quorum Witness starting on page 22 in the  "HPE 3PAR Operating System Upgrade Planning Guide": http://h20564.www2.hpe.com/hpsc/doc/public/display?docId=emr_na-c02660486

2. Do a Manual Transparent Switchover of all remote copy groups

Find the steps in the "Remote Copy User Guide" here  http://h20565.www2.hpe.com/hpsc/doc/public/display?docId=emr_na-c03618143

3. After checking that all your remote copy groups are on the surviving site you can safely bring down your freed-up site

 

Cheers

Pete

ArunVC
Visitor

Re: 3PAR Peer Persistence with VMware Stretch Cluster

Hello,

To do a manual failover you dont need to bring down the quorum server.

One thing to check before a full failover is that

1) all your rc groups are completed synced

2) The best method for setting up replication is to have different rc groups based on cluser or app/db relation. Becasue if all the luns are put into a single group for replciation, during a failover it can cause a huge load and whcih may casue undesirable results