According to this document, the following DISK configuration changes should be made when connecting vSphere 4/4.1 to an HP EVA active/active array.
1. Change the default SATP option from VMW_SATP_EVA to VMW_SATP_ALUA.
2. Change the default PSP option for VMW_SATP_ALUA to VMW_PSP_RR.
3. Configure round robin advanced parameters to IOPS=1.
What's not clear if I should also make these changes to the HP Fibre Channel RAID Ctlr listed as LUN0 in the storage adapters page (type = array controller).
Any advice?
The last time I configured ESX on an EVA 4100 (Firmware 6.200), I just followed the steps on page 18 on the CLI. After a reboot of the ESX host the settings were applied correctly.
VMW_SATP_ALUA and VMW_PSP_RR for the disks
André
I just got done setting up a EVA 4400 with vSphere 4.1 and followed the exact document. I don't remember seeing anything about STEP 1. I only peformed STEP 2 & 3.
Step 2 just makes it so that each new lun that is added to a host defaults to the RR policy.
Step 3 changes the IOPS for the LUN to 1 on all the LUNs except LUN 0. I believe LUN 0 is a system lun that you don't want to mess with.
These are the commands I used:
Script to get existing lun configuration.
for i in `ls /vmfs/devices/disks/ | grep ^naa.600`; do esxcli nmp roundrobin getconfig --device $i; done;
Script to change existing lun configuration from x IOPs to 1 IOP on all LUNs.
for i in `ls /vmfs/devices/disks/ | grep ^naa.600`; do esxcli nmp roundrobin setconfig --type "iops" --iops=1 --device $i; done;
Changes the default PSP policy applied to NEW luns. Reboot the ESX host to have it change it automatically on existing LUNS otherwise you have to do it manually one by one.
esxcli nmp satp setdefaultpsp –satp VMW_SATP_ALUA –psp VMW_PSP_RR
When you run these scripts, make sure you don't have any other SAN luns attached to your ESX hosts. The EVA should be the only one presenting luns to the hosts.
1, correct, step 1 not needed, vSphere 4.1 should use VMW_SATP_ALUA itself
2, VMW_SATP_EVA to manage raid controller lun 0 also correct, no need to change to VMW_SATP_ALUA
binoche, VMware VCP, Cisco CCNA
1, correct, step 1 not needed, vSphere 4.1 should use VMW_SATP_ALUA itself
2, VMW_SATP_EVA to manage raid controller lun 0 also correct, no need to change to VMW_SATP_ALUA
binoche, VMware VCP, Cisco CCNA
Edit: I was wrong, our disks are defaulting to ALUA.
According to the best practice document for the EVA you should not use the VMW_SATP_EVA.
VMW_SATP_DEFAULT_AP is the SATP that active/passive EVAs will use. However, active/passive EVAs are not supported with ESX 4.0 thus using this plugin to enable ESX 4.0 connection to an active/passive EVA for production is not recommended. VMW_SATP_EVA has no useful application. It is an SATP for active/active EVA models, which have the TPGS (ALUA compliance) option turned off. By default, all active/active EVAs have the TPGS option turned on and this option is not user configurable. Thus no EVA arrays currently use this specific SATP (VMW_SATP_EVA).
could you please post the below info? thanks
1, vmkernel messages
2, esxcli nmp device list; esxcli corestorage claimrule list; esxcli nmp satp list;esxcli nmp satp listrules
binoche, VMware VCP, Cisco CCNA
could you please post the below info? thanks
1, vmkernel messages
2, esxcli nmp device list; esxcli corestorage claimrule list; esxcli nmp satp list;esxcli nmp satp listrules
binoche, VMware VCP, Cisco CCNA
I was wrong, the disks are showing up as AULA.
Brian Olsen wrote:
According to this document, the following DISK configuration changes should be made when connecting vSphere 4/4.1 to an HP EVA active/active array.
1. Change the default SATP option from VMW_SATP_EVA to VMW_SATP_ALUA.
2. Change the default PSP option for VMW_SATP_ALUA to VMW_PSP_RR.
3. Configure round robin advanced parameters to IOPS=1.
What's not clear if I should also make these changes to the HP Fibre Channel RAID Ctlr listed as LUN0 in the storage adapters page (type = array controller).
Any advice?
So did you ever figure out if you needed to make changes to the controller lun0? I am adding a new 4.1 host to an existing cluster with an HP EVA 4400 array. Wondering the same thing. The other hosts in my cluster which predate me at this job have the array controllers set to VMW_SATP_DEFAULT_AA. My new host has the array controllers set to VMW_SATP_EVA but I would like to keep these consistent.
If I do go and change it what's the syntax of changing SATP for individual devices/paths?
There is no need to make any changes to LUN0. You only make the changes to the datastore LUNs.
Stating the obvious of course but using perfmon to view the host port stats will show how multi-pathy things are.
Also as per attached graph of read activity on my EVA4400, if IOPS=1 then should only see 2 lines as numbers of reads each port will be the same - and also the load variance across the controller ports with how the LUNS have been presented.
They should all be set to ALUA. Just do a:
esxcli nmp satp setdefaultpsp --satp VMW_SATP_ALUA --psp VMW_PSP_RR
and reboot the host so that all LUNs that were already presented are reset to ALUA/RR. Also if you want to change the number of IOps path should be swapped use the following to do it for every LUN:
for i in `ls /vmfs/devices/disks/ | grep naa.600` ; do esxcli nmp roundrobin setconfig --type "iops" --iops=1 --device $i ;done
Duncan (VCDX)
Available now on Amazon: vSphere 4.1 HA and DRS technical deepdive
Thanks for the replies everyone. I already ran esxcli nmp satp setdefaultpsp –satp VMW_SATP_ALUA --psp VMW_PSP_RR
I am just now curious as to what the difference is between having an EVA controller set to VMW_SATP_DEFAULT_AA versus VMW_SATP_EVA? If there is no difference that's fine but I would still like to know how to change this setting on just the controllers in case I wanted to make it consistent on all my hosts. Like I said earlier in my cluster of 12 hosts 11 of them have their EVA controllers set to VMW_SATP_DEFAULT_AA and one has it set to VMW_SATP_DEFAULT_AA.