EVA 4400: Cannot move LUNs to Controller 2

[pinkerton]

Hi,

when setting the preferred path/mode to Path B for a given LUN, both Command View and EVAperf show Controller 2 as Managing Controller for that LUN. However, after about one minute this automatically reverts back to Controller 1.

This is the case with both Windows Server 2008 Operating systems (ALB == disabled) and ESX 4.1 (Fixed Path to Controller 2).

This even happens when creating a new LUN from scratch with Controller 2 as managing controller - it automatically changes to Controller 1 even if the LUN is unpresented and there is no IO at all.

The EVA is is running 09522000. Is this a know problem? I haven't found anything regarding this in the release notes of Hi,

when setting the preferred path/mode to Path B for a given LUN, both Command View and EVAperf show Controller 2 as Managing Controller for that LUN. However, after about one minute this automatically reverts back to Controller 1.

This is the case with both Windows Server 2008 Operating systems (ALB == disabled) and ESX 4.1 (Fixed Path to Controller 2).

This even happens when creating a new LUN from scratch with Controller 2 as managing controller - it automatically changes to Controller 1 even if the LUN is unpresented and there is no IO at all.

The EVA is is running 09522000. Is this a know problem? I haven't found anything regarding this in the release notes of 09534000.

[pinkerton]

Hmm. Now it seems to work...

Though I have written above that the PSP in ESX 4.1 was set to "Fixed" with a preferred path to Controller 2, I first changed this today. Before this morning I was using Round Robin and ESX automatically chose 2 paths to Controller 1 as preferred paths.

Now that "Fixed" is configured for some hours, EVAperf shows that all ESX LUNs automatically changed to Controller 2!

The Windows 2008 LUNs also changed to the controller that was configured in the preferred path/mode option. I.e. the LUNs which were configured with path B moved from Controller 1 to Controller 2 (and I haven't changed anything on the Windows Servers!)

In addition to that the configuration doesn't revert back when changed anymore.

Currently I don't understand this. But it seem related to the massive amount of
"03260027 - An HSV300 controller has changed Battery Cache policy" messages that appeared in the controller event log until a few hours ago.

[pinkerton]

Better don't cheer to soon. Just checked EVAperf again and all LUNs have moved back to Controller 1 again - right after the controller (1) reported "An HSV300 controller has changed Battery Cache policy" again.

...five minutes later...

One LUN has automatically reverted back to controller 2 now. Maybe it takes some time until the others follow.

[Torsten.]

This could be a reason. The ownership of a LUN goes away from a controller with a bad battery.

[pinkerton]

The strange thing however is that it is controller 1 that is reporting the battery problems. I just checked the enclosure and found out that the battery pack of controller 1 is slowly flashing green, while the battery of controller 2 consantly shows green.

[Torsten.]

Somebody should have a look into the logs to track this down.

[pinkerton]

Might it be worth a try to reseat the battery pack or reboot the controllers?

[Daniel_279 · Accepted solution]

Two fixes in XCS 0953400 Release notes specific to EVA4400

Added a verification step to ensure that an accurate cache battery state is returned for HSV300 controllers to avoid falsely triggered cache policy changes.

On the EVA4400, false reporting of the battery holdup time caused a change to the LUN cache policy.

Restart effected controller will resolve it. Updating XCS code will prevent it

Also search google for "Configuration best practice for HP StorageWorks Enterprise Virtual Array (EVA) family and VMware vSphere4"

[pinkerton]

Thanks, I missed the note about the verification step regarding the battery state! Guess this will solve it then.

[pinkerton]

Just to let you know: I've rebooted controller 2 and for now there are no messages regarding the battery policy anymore. In addition to that the status indicator of battery 1 (controller 2) doesn't flash green anymore but shows constant green now.

At least I know how this can be solved until we can uprade to the new XCS version.

BTW: When I first tried to reboot the contoller using WOCP it didn't come back online again. In order to get the controller working again I had to reseat it!