Unsolved
This post is more than 5 years old
1 Rookie
•
33 Posts
0
1189
August 18th, 2016 05:00
Is this a best practice for rp4vm iscsi? vmware say no.
In a best practice doc from inside EMC, I find below slide:
I think it means that vnic of vrpa iscsi can use the same vswitch and NIC as the ESXi host ISCSI adapter, even if this is not recommended. But in a customer environment, user use this configuration, and they meet a serious performance issue. VMWARE engineer checked the environment, and say that the root cause is vRPA'S ISCSI VNIC share the same NIC with ESXi ISCSI ADAPTER, and this is prohibited. Anyone has esperience about this?
No Events found!
fadliz
153 Posts
0
August 18th, 2016 07:00
Hi there,
Can you give us more details on the following:
VMware recommends using the same broadcast domain b/w the vRPA iSCSI interfaces and the VMKs being bound to the adapter.
Also, is there a SR number w/ EMC?
I am not sure where this slide came from, do you have link?
Feel free to contact me offline if you would like to share the info above.
Thanks, Zahid
liuchuanwen
1 Rookie
•
33 Posts
0
August 18th, 2016 21:00
hi Zahid,
I don't have detail IP configuration information, because it's not implemented by me. But I know this is a pure iscsi environment, no FC SAN, the product VM's vmdk also use the same ISCSI adapter, I guess maybe this is a problem.
The link of the document I refered to is : https://inside.emc.com/message/664537#664537
There is no SR in EMC, user met the issue and vmware drew a conclusion like that, and then user stop all CGs , and delete all VRPA VMs immediately, and ask us to draw a conclusion whether adding new NICs to separate vRPA's VNIC from ESXi ISCSI adapter can solve this performance issue, and after that we can redeploy the RP4VM, so we don't have any logs can provide to support to analyze.
Thanks.
fadliz
153 Posts
0
August 19th, 2016 02:00
Hi, Have they disabled delayed Ack? Check VMware Kb1002598 we recommend disabling it especially in an iscsi env. . And if the BE storage is iscsi, it might be a resource contention hence the recommendation.
Thanks! I will take a look at the doc.
And if possible please have then open an SR with EMC if they can redeploy.
Sent from my iPhone