Lefthand P4300G2 slow performance after upgrade

[Mick Negendahl]

Hello,

I recently upgraded our Lefthand P4300 to Lefthand OS 12, since the upgrade i have experinced a performnce drop - after much checking the fourm i found a lot of people mention that upgrading the memory seems to help, I have checked the memory utilization on my Lefthand, and it rarely goes above 50% for any of my 4 nodes.

I experience the problem with my VMware environment and my Veeam Backup server that has direct SAN access, due to both of them having the same kind of very slow transfer from the SAN i expect the SAN cluster is the fault.

I tried pretty much everything i could try to improve performance, except swapping the memory - would it help?

here is my set up (and preciously worked perfect):

4 Nodes, clustered

Every node has 4GB of memory

Network raid 5 across all nodes

ISCSI connection to all servers

Data spread across 6 volumes

Seperated VLAN for storage traffic

2 NIC pr. node teamed and using load balancing connected to storage VLAN

Servers set to use load balacing in CMC

Flow control enabled all around

Jumbo frames enables all around

The CMC console is not reporting any kind of problem, but since the upgrade i get a latancy warning from vSphere every 1-10mins. 

Thanks!

[Mick Negendahl]

The CMC monitor seems to point to massive write lantecy, it maxed out at 366ms and avrage is 38ms. 

Seems a bit high compared to the read lantecy.

Any help is much appreciated.

[oikjn]

I would verify that jumbo frames actually works by sending a jumbo packet ping to make sure it actually gets through.

past that, if you are using network raid 5 then slow write performance is expected.  HP only suggests that for read-mostly archive data like ISO storage archives.  NR10 is the defacto only way to go for network raid.

[Mick Negendahl]

Thank you very much for your reply.

I have now verified that jumboframes is working.

I have opened a support case at HP on the matter, hopefully they can help me out with the problem.

[oikjn]

are these LUNS setup as NR5?  That can definitely cause performance issues, but if you were doing that before v12 I would definitely get support involved as there could be something deeper going on that they will want to fix.