P2000 G3 MPIO not working as expected.

[Lianlk]

I have a P2000 G3 with two controllers and ports A1,2,3,4 on top and B1,2,3,4 on bottom.

On Server 2012, I configure 2 nic ports and connect directly with 2 cross-cable to port A1 and port B1.

I try round robin but I see only one LED flashing on A1. B1 is idle. 

If I try to plug out A1 to test failover, it either take a long time or OS will errors out.

Help?

[Torsten.]

Remember the array is an ALUA device, so the owner controller will do the work. But failover should work, maybe you need to double check the servers settings and configuration.

[Lianlk]

You mean beacuse this P2000 is ALUA, I cannot use Microsoft MPIO to setup for double performance from dual link like now -- nic1 to A1, nic2 to B1?

Then how to achieve what i want?   I using iSCSI, dual 1Gb link to p2000 controllers port A1 & B1?

[Torsten.]

With ALUA you can have active multipathing between the ports of the LUN owning controller.

[Lianlk]

Thanks for the effort but afraid I don't follow what you are trying to say.

This P2000 is turning out to be a bad purchase.  

We should have looked at other manufacturers SAN box that can at least work with Microsoft MPIO as expected.

[Torsten.]

ALUA is an industry standard. Many arrays work based on this.

You should make sure you use the latest drivers and firmware (at least TS251P006), because there were some changes regarding win 2012 recently.

From the notes:

The following features were added or enhanced in controller firmware TS250R021:

• Added support for Microsoft Windows Server 2012.

[Manfri]

If you have 2 controller a single set of disk ( vdisk) is managed from a single controller.

And the two controller are linked internallyand provide a second path, with an intermediate hop

For every vdisk managed by controller A ,with 2 nic working, you'll have 2 path, 1 optimized (the one connected to controller A) and 1 not optimized ( the one connected to controller B ).

and viceversa

For every vdisk managed by controller B,with 2 nic working, you'll have 2 path, 1 optimized (the one connected  to controller B ) and 1 not optimized ( the one connected to controller A ).

Which kind of Round Robin policy you use? as i remember the default round robin policy is

- RoundRobinWithSubset – In configurations where some I/O paths are faster than others, this
variant of the RoundRobin policy does not distribute I/O to the slower paths, as long as at least one of
the faster paths is available.

taken from HP StorageWorks MSA System MPIO DSM Installation Guide

So use only the only optimized path to the controller who manage the vdisk... 

This is an high available configuration and to use both path you must use this policy

- RoundRobin – All paths are active paths. They are used for sending I/O in a round-robin fashion.

or have a fast but NOT HA solution where each server is connected to a single controller, so you have 2 optimized path and 2 unoptimized path

In fact i try to distribute the disk between the two controller so both controller CPU are used (and i gain some IOPS especially in raid 5) , both path are used ( 1 for vdisk on controller A and one for vdisk controller B ) 

PS: never use LUN 0. it's used for some internals ( VSS snapshot for example ) and there are configuration where this run interference and strange problems, number LUN 1 and up and these problem disappear )

[Torsten.]

>> With 2 nic working you'll have 4 path, 1 to controller A optimized ( direct ) 1 to controller B optimize, 1 to controller a unoptimized ( indirect ) and 1 to controller B unoptimized

Sure?
I think all the paths to the owner controller are optimized only (e.g. A1, A2, A3, A4); the B - paths are not optimized, because controller B has to "hand over" the operations to the owner controller A.

[Manfri]

the phone interruptions generate monster ... i've corrected the post