single path problem with EVA

[Hanh_3]

Hi,
I have 10 single-path Windows servers and an EVA5000 v3020.We do not have Secure Path. I have zoned 5 servers with Controller A and the other 5 servers with Controller B. Each server can see only one LUN. But when I copy data to LUN, there is problem with servers which is assigned to Controller A. Shut down controller A and it's ok.
So I can not load balance vdisks between two controllers. Can anybody help??

1. A vdisk will be taken by which controller when it is created first?
2. If a vdisk is taken by Controller A, but the server that the LUN is presented is zoned with Controller B, then LUN can change to Controller B ??

Thanks
Hanh

[Uwe Zessin]

1) It depends on the preferred path setting. Choose 'no preference' and the EVA will distribute round robin.

2) Yes, but the server needs to tell the EVA to fail over the virtual disk to controller B.


I suggest you use explicit preferred path settings with the failback option.

[Richard Perez]

Hanh
have you tried MPIO? With zoning you are depending on a single controller to access the LUN, but with MPIO you will be able to access the LUN by either controller

Regards

[Hanh_3]

I have configured vdisk with either "No preference" or "Path A failover" (server zoned with Controller A) but still error.
When copy data, server's port on switch is offline, online and repeat forever. Port error count number increses constantaneously.

There are 2 controller events:
- An excessive number of link errors were detected on a HSV110 controller's Fibre Channel port
- A non-data exchange sent to a physical disk drive has timed out.

Try to unpresent lun but fail and EVA change to "Unknown Storage System".

It happen with every server that I present LUN to. So I don't think this is the error of switch port or cable or HBA. It may be the multipath error.

At first I didn't use MPIO because some one in this forum said that it not support Business Copy. But now I think I should try using MPIO. I will post the result later.

Uwe: Have you ever try to configure single path server with single path zoning and vdisk set to "Path A/B failover" or "Path A/B failover/failback" and it works properly ? I don't think it support those options without multipath software. As you said "server net to tell the EVA to fail over the virtual disk to controller B" , but how server tell EVA if there is no multipath software.


Thanks

[Hanh_3]

I attach here the 2 controller events for more details.

[Uwe Zessin]

If your fibre channel switch ports go offline, then you have another problem that is unrelated to single pathing, but affects all your experiments. You must fix that problem first.


From you descriptions it is not clear to me if the controller events deal with the host port, but it sounds like you have problems on the back-end, too. Sorry, I can't process your '.rar' archive.

If the EVA changes to "Unknown Storage System" it means that the management agent has lost contact with the EVA. What type of management server are you using? If it is not an SMA, have you made sure you are using the correct driver revisions?


You didn't talk about BC in the first place. I had a chat with HP presales last week and got told that the Windows MPIO is now supported, but I am waiting for a compatbility list I can check.

About your last paragraph: my point was that the driver in the server does not know how to do a failover *because* there is no multipath software. So you need to tell the EVA to keep the virtual disk on one controller.

[Hanh_3]

Uwe: Sorry, I misunderstand your reply :-)

I reattach the event with port error here.
"Severity: Normal -- informational in nature. An excessive number of link errors were detected on a HSV110 controller's Fibre Channel port. The identity of the affected Fibre Channel port is contained in the cerp_id field. The HSV110 controller internal Fibre Channel port number of the affected Fibre Channel port is contained in the port field. This is an informational event triggered by the occurrence of an excessive number of Tachyon chip link status errors detected within a particular link status error type. The number of occurrences of each link status error type is contained in a separate type-specific field (e.g., loss_of_signal, bad_rx_char, etc.). The HSV110 controller checks each of its Fibre Channel port's for excessive link status errors periodically.
...
cerp_id - HSV110 controller enclosure rear panel Fibre Channel port: MP
...
port - HSV110 controller internal Fibre Channel port number: 4 [0x04]
... "


I guess MP is Mirror Port.
What about "controller internal FC port number 4" ??

Yesterday I did some tests, and here is result:
- When server is zoned with Controller B, it will work well with vdisk set to either "Path A failover/failback" or "Path B failover/failback"
- But when server is zoned with Controller A, it cannot work well with vdisk, and the controller errors occur.
- Test with only Controller A (Controller B is shut down), every thing ok.

So if there is problem with mirror port, why it still work well when zoned with Controller B?
Or maybe I'm wrong with mirror port. So what is "MP" ??

Thanks

[Hanh_3]

Hi,
I have changed the copper cable between two mirror port to an optical cable and it works :-).
So there is nothing error with single path.
Thank you guys for replying.
I'm going to close this thread.

Regards,
Hanh

[Hanh_3]

See above.