MSA1500 redundancy

[James Mercer]

Hi,

We have a dual controller MSA1500cs. Each controller has a fibre channel I/O connector behind it. Each FC I/O is connected to one san switch (FC 8 port 4Gb/s from HP). There is no zoning on the switch.

We've then connected servers with single FC connections to the switch.

The MSA1500 reports that the controllers are in Active/Standby mode. The right (facing front of MSA) controller is showing active and the left controller is showing standby.

If we disconnect the fibre channel link to the primary controller, we lose ALL communication with the MSA - no disk/lun access and no ACU access despite the fact that we still have one working FC connection to the MSA and the FC connections to each server. We've swapped controllers, FC cables, FC I/O cards around on the MSA.

I've logged this with HP and they initially changed the chassis of the MSA. However, we still have this problem. Our firmware was 5.02 and we tried 5.10, but nothing has changed/improved.

Does the second FC connector work at all in these things? Is the redundant controller able to do anything? Should we move our second FC I/O connector to be behind the active controller?

Thanks

[Torsten.]

Do you have path management software like HP Secure Path installed?

Quote: "NOTE: Redundant configurations require two Host Bus Adapters per server, an additional controller and fibre channel I/O module, redundant switches and cables, and Secure Path (active passive configuration) software or Industry Standard failover depending on the operating system for each server. "

source: http://h18000.www1.hp.com/products/quickspecs/11945_div/11945_div.html

[James Mercer]

Thanks - I've been reading that passage many times today. We realized that redundant paths require redundancy throughout, but we expected the MSA to provide redundancy within.

In otherwords, regardless of whether we had single or multi-pathed hosts, we expected the MSA to be able to fail over to the other controller. That way if one controller failed the other would take over. I'm (almost) certain this is the case in large HP SANs/controllers. It appears that HP relies on external software and paths for the MSA instead of internal busses and failover. Is that true?

[John Kufrovich]

James
Torsten is correct. You need either Secure Path or MPIO for Active/Standby.

We recently released A/A FW for the 1500. Ver 6.86. You will need to download the MSA Support Cd 7.57. It will have all the necessary items. Also, Secure Path isn't supported but Full feature MPIO is.

jk

[James Mercer]

Ok, so redundancy 'within' the controller is only available using SecurePath or MPIO. Yes/no?

We have a mixed environment including Netware. This seems to prevent us from using Active/Active configurations. (HP recommends that we use firmware 4.98 or lower, but only supplies 5.02 and higher.) Is there anywhere I can get appropriate firmware for use with Netware?

I understand MPIO basic is free. Is MPIO Full also free (only usable with Active/Active firmware)?

[John Kufrovich]

Yes,

Basic and Full feature MPIO are free.

Let me ping a couple of individuals to find out what the status of Netware MSA FW support.

[Torsten.]

generally speaking, if an array has dual controllers, the controllers *must* cummunicate with each other to prevent things like a "split brain". This is true for *all* arrays. Some arrays have active/passive controllers (passive become active, if the active controller fails), some have active/active. Most OS needs special software to be able to talk to an array with dual pathes. In case of active/passive, the OS has to switch to the redundant path. In case of active/active, the OS has to know, that the "2 different" devices seen are in fact only one with dual pathes. A nice feature in this case can be load balancing.
In fact, if your active controller fails, the passive should become active.

If you only disconnect the fibre, the controller is still working, no reason to fail-over from his point of view.

Now your host have to switch to the other path and force the controllers to switch.

This is the concept in general.

[James Mercer]

John - if you can find firmware 4.96 that would be great. We have this on one of our MSAs and it seems to be reporting everything is ok. (In Insight manager it is reporting both controllers are ok, while the MSA with 5.10 shows one controller as red in Insight Manager.)

Terton - Thanks for the details. Yes, if the FC is removed from the primary/active controller fail over wouldn't happen without path software.
However, we have also tested this by removing the active controller. The standby controller tries to become active and then reports "43 Redundancy failed hardware failure". We were trying to simulate loss of the active controller. This test is not supported by HP; is it a valid test of redundancy? Is there another way to test the controller redundancy? Despite the chassis change, the test still fails and the standby controller will not become active.
Do the controllers fail over to each other at all without external software?

Thanks

[James Mercer]

Sorry Torsten, I mis-spelled your name in my reply. Many apologies...

[Torsten.]

IMHO the removal of the fibre cable is not a reason for the controllers to switch. From their point of view everything is fine. They can talk to each other and are able to see all disks.

As you mentioned, to force the fail-over is part of the software on the hosts. You will need the software, there is no way around.

Even if you have an OS that supports and expect dual active/active pathes, you need special software for active/passiv arrays, because every access to the passive ctrl would cause the controllers to switch. This would result in a endless fail-over from one controller to another.

Your test is realistic, but it is not recommended, because you decrease the redundancy without emergency and you can damage the parts by accident. You do this at your own risk.

But you have seen that the fail-over is working ;-)