SecurePath Auto-Restore feature question

[Informatik_1]

Hi Guys

We tested the redundancy of our SAN (EVA3000, SecurePath on every Server, 2 HBA's on every server, 2 switches). We turned of one switch, but some of the servers lost their EVA LUN's. The zoning is correct, but I figured out, that the Auto-restore feature is set off on every server.
I think the feature has to be set on, otherway SecurePath doesn't repair a link.
Could someone please explain that?
And what is the improvement if I enable the Load Balance feature too?

Thank you very much

Marc

[Uwe Zessin]

If you turn off one switch and some servers loose access to their storage, then you have a configuration problem - did you make sure that all paths work properly? How do you know that the zoning is correct? I have given up counting the situations where I thought my zoning was correct - one of my standard errors is that I create a zone, but somehow forget to add it to the configuration...

It looks like the comments for Auto-restore are misleading, because in the GUI it is called Auto-Failback. That is not always such a great idea, because it requires the controller to flush the LUN cache and transfer the ownership to the other controller - that costs time.

Load balancing means that you can do I/Os to all ports of the controller that currently owns the LUN. Using a 'standard cabling' it also means that data goes over both fibre channel adapters. It is turned off by default, because there can be issues with SCSI device reservations in cluster environments - the documentation has a table that explains what is supported in which situation.

[Informatik_1]

Thank you for the quick answer.
So turning off the Auto-Restore feature doesn't prevent our SAN environment from being redundant? Is that correct?
The initial zoning has been done by HP.
Attached you find our zoning sheet, do you see any misconfiguration?

Thanks, Marc

[Vladimir Vybiral_1]

No, having autorestore set to off does not compromise the availability. It is recommended to leave it off - as if You would have a link that is "flaky" for data transfer as a preferred one, the LUN would be travelling all the time between links as the flaky one would go on and off and on and off... With Aoterestore off, You have to go to the mgmt and manually switch the path back to the path that has failed and is repaired.

BR Vladimir

[Informatik_1]

Thanks Vladimir

So SecurePath automatically detects a failed path and switches to a new one if one is available. After the failed link is available I have to manually repair the link with spmgr.
So the only thing that could be wrong is the zoning, nothing else?

[Uwe Zessin]

Marc,
that is the zoning _plan_ - it does not show how the zoning is really set up. A 'cfgShow' output from the switches would reveal this.

I think the 'repair' is misleading - it will detect when the path is working again, but the default is not to automatically switch back for reason I and Vladimir have already explained.

Of course there are many other things that _could_ be wrong.
- missing links
- not all adapter WWNs are assigned to the hosts on the EVA
- ...

Before diving into too much details I would check on the host for missing / failed paths. Look for the 'spmgr' program on your server(s) and check the output of 'spmgr display'.

[Informatik_1]

Thank you very much, the zoning on the switches reflects the attached excel sheet. Every Server can see his LUN's over the right paths.
So I have to test the whole environment during a weekend and try to find out if all servers loosing their LUN's after switching of a switch or not.

Marc

[Alzhy]

Show us a "spmgr display" output on servers which "lost" their LUNs during your tests as it will show your current config.

If may be possible that you're HSV to switches connections are not correct. Our EVA's and SAN are structured this way:

We've 2 independent SANS - SAN A nd SAN B. Each SAN has its own Core-Switch - Switch A and Switch B. Each HSV110 (EVA Controller) connects to to each Switch. And on each server, one HBA connects to SAN A and the other to SAN B. This configuration allows for shutdown and maintenance of either SAN.

On the UNIX side (we've an all UNIX connectvity to our EVA SAN - a mix of Sun and HPs) - we do load balancing ("spmgr set -b on") and auto-restore ("spmgr set -a on"). We disable PATH/COntroller preferencing on each LUN from the EVA and just let SecurePath manage the load and controller balancing...