VMware Cloud Community
dahvaio
Contributor
Contributor

Emulex Queue Depth - lpfc# - What should # be?

I need to change the Queue Depth to 16; however, I am not sure the exact command to use:

My ESX host has 2 Emulex HBA's, when I viewing /proc/scsci/lpfc - I get the following:

**************************************

# ls

2 3

**************************************

If I use the following command to change the Queue Depth, what is the correct lpfc# to use:

esxcfg-module -s "lpfc#_lun_queue_depth=16 lpfc#_lun_queue_depth=16" lpfc_740

lpfc0 and lpfc1

esxcfg-module -s "lpfc0_lun_queue_depth=16 lpfc1_lun_queue_depth=16" lpfc_740

or

lpfc2 and lpfc3

esxcfg-module -s "lpfc2_lun_queue_depth=16 lpfc3_lun_queue_depth=16" lpfc_740

Does the lpfc# correspond to the SCSI device or the HBA number?

Thanks...

Reply
0 Kudos
6 Replies
RParker
Immortal
Immortal

nothing, leave it alone. Earlier builds of ESX needed a work around (due to software restrictions and driver bugs) that were causing a conflict in the queue depth, no longer an issue.

The default is 64, but in any case you should not touch it.

Reply
0 Kudos
dahvaio
Contributor
Contributor

I use a Hitachi AMS2500 and the Vendor has told us that we had to set the Queue Depth to 16. At first they told me me it had to be set to 4.

Reply
0 Kudos
RParker
Immortal
Immortal

Queue depth means that IF you have outstanding SCSI commands, the lower the depth, the commands will be dropped once the queue is reached.

So hitachi should not be making recommendations for stuff that can change and for products that are not standard OS (ESX isn't a simple OS). My guess is you didn't tell Hitachi you were using this for ESX.

I still say don't change the default.

Reply
0 Kudos
dahvaio
Contributor
Contributor

RParker,

I understand where your coming from; however, I am at a point that I do not want to argue with Hitachi. They know exactly what this is for and they are stating it is THEIR BEST PRACTICE. This is coming from their Top Hitachi VMware SAN Engineer. Not sure if they are drawing at straws but who knows...

If you know the answer to what I am looking for then please answer it...

Should it be "lpfc2 and lpfc3" or "lpfc0 and lpfc1"

Thanks...

Reply
0 Kudos
McSatty
Contributor
Contributor

I'm traveling to PMC Montreal on project related work. My response to emails will be delayed

Thx, Satheesh

Sr Enterprise Storage Management Expert

PMC-Sierra Inc

Burnaby, BC

Cell: 604.788.0119

Email: smc@pmc-sierra.com<mailto:smc@pmc-sierra.com

Reply
0 Kudos
PacketRacer
Enthusiast
Enthusiast

It's lpfc2 and lpfc3 for you... But you should reboot, if possible, and verify that after the reboot you are still seeing lpfc2 and lpfc3. Those number could change, if I'm not mistaken.

Just curious... what is the bigger problem? How did you get into a situation where Hitachi is asking you to change the queue depth?

I'm running an AMS 2500, too, and haven't had issue. I did not have to change the queue depth, but I only have 8 hosts with 2 FC ports each. Hitachi supports a max queue depth of 512 per controller FC port, and recommends that you calculate the queue depth based on active paths per port. So if you have 16 active paths per AMS FC port, then you have to set your HBA queue depth to 32.

512 max depth per AMS port / 16 path = 32 per path

If you have a lot of hosts and/or a lot of HBAs per host, you may have to tune the queue depth, per Hitachi's official recommendation.

Reply
0 Kudos