Array Performance and Data Protection
cancel
Showing results for 
Search instead for 
Did you mean: 

Peer Persistance switchover, iscsi timeout

 
SOLVED
Go to solution
Advisor

Peer Persistance switchover, iscsi timeout

Hi,

I am novice in Nimble solution storage.

I want to know if my configuration is correct for have a peer persitance and automatic switch over.

We have two AF20 with 2 10Gbit/s/ sfp+ ports per controller. ie. 4 per array for Data ang group trafic

and 2 10Gbit/s BaseT per controller for management.

The 1st port SFP+ of each controler reserved for iSCSI  - single subnet

and 2nd SFP+ is reseved for Group synch - single subnet

The arrays are installed in two sites and linked by 20gbits/s fibre

 

Clip2net_201127161250.png

When I test the ASO, the VMs is locked for 30-50 seconds before Nimble can deliver IOs, is it a normal situation or not?

3 REPLIES 3
HPE Blogger

Re: Peer Persistance switchover, iscsi timeout

Hello,

when performing some kind of failover across sites, it's commonplace to have an IO pause in order to ensure that:

a) data doesn't get corrupted
b) you don't experience "split brain" with both sites thinking they're active due to a temporary outage.
c) your VMs don't get destroyed or corrupted

Therefore 30-50 seconds of IO pause (NOT timeout) is expected and normal.

However - i must warn you on your network configuration. Only having a single 10Gb port for your iSCSI data most likely is going to cause problems - as you have no resilliency in your network should something happen on your switch. And Peer Persistence will not failover should a switch fail. Your current network setup is deemed to be non-recommended, and will most likely cause issues somewhere down the line.

Instead, run all four ports across both arrays trunked with both group and iSCSI data on different VLANs across those ports. this will give you the resilliency you need.

Nick Dyer
Global Storage Field CTO & Evangelist

twitter: @nick_dyer_
Advisor

Re: Peer Persistance switchover, iscsi timeout

Hi Nick,

 

thanck you for response.

 

Ok for "IO pause".

For "teaming" the iSCSI and Group interfaces, can i do it without redeploing the arrays?

Just add VLAN tag for 10Gbits interfaces at the same time as changing switch port configuration?

I can stop the host's IO, the arrays is not in production yet.

 

regards

HPE Blogger
Solution

Re: Peer Persistance switchover, iscsi timeout

Yessir, you can jump into the network configuration on the system and edit the setup to create VLANs on the NICs very easily. What's more, you can save the configuration as a draft - to ensure that maybe some networking folks check it over before you commit. Should you then commit and something not work - not a problem, you can restore the previous network configuration back.

Please feel free to give Nimble Support a call whilst you're doing the changes - they'll be able to walk you through it too.

Thank you for your custom and trust

Nick Dyer
Global Storage Field CTO & Evangelist

twitter: @nick_dyer_