Host Errors in 3PAR V400

[kk2491]

Hi Guys,

Good Day.

Before explaining the issue i will give a brief background.

We have 3PAR V400 with AO and QOS enabled, the LUNs from this 3PAR array is given to VMware infrastructure.
Recently VMware team has noticed that some disconnection errors in VMware end and also reported that the VMs are getting affected as there is some Read only issue.

Errors in VMware

Capture.JPG (27.48 KiB) Viewed 15268 times

Later I checked in 3PAR array (in debug log) and found many host errors in the system.

Can anybody please let me know the below
1. What causing this Host errors ?
2. What is the impact of these host errors ? does this cause outage to servers ?
3. How can we prevent this from happening again ?

Thank you. Awaiting replies ...

Regards,
KK

[mujzeptu]

I have had similar issues in the past on my V400 with I believe the 3.2.2 firmware? Anyway we fought and fought the issue and as I recall it was a HBA firmware/driver issue on the host.

I would make sure your hosts have the latest-greatest firmware and drivers and ESXi image from the vendor.

[kk2491]

Thanks for the reply.

I think the hosts HBA version are the compatible ones. Let me check and update.

Regards,
KK

[MammaGutt]

I've at least seen this log entry on:
Bad FC cable on 3PAR
Bad FC cable on host
ATS issues
Bad HBA drivers in Vmware
Buggy SAN switch FW

And that doesn't put AO and QoS into the mix.... But since you are mentioning those, are you only seeing the log entries when AO is running and when QoS is active? Remember that QoS does throttle IO so it could also throttle heartbeats/keepalive messages from Vmware if it is kicking in.

So I would suggest to limit the scope of your issue and start eliminating everything you can... AO and QoS can easily be disabled for a period to see if it matters.

[Richard Siemers]

Tell us about your zoning strategy. Does each ESX host have 2 HBA ports? How many 3PAR ports is each host HBA zoned to?

Are the errors you see in the 3PAR log talking to a LUN 256? What version of ESX is this?

There was a known issue when ESX upped the LUN limit from 256 (0-255) to 512... some vendors were using LUN 256 for in-band management, VVOLs, etc. So when ESX was upgraded it started seeing a new LUN that was not storage and sending it status and inquiry requests.