EMC PowerPath Install

[Youlette Etienne_2]

Hello Everyone,

Last week, we migrated our SAN from a cx600 to a cx700, with EMC onsite. I spent quite a significant amount of time performing the vgexport/vgimport commands for 7 servers since the device path names changed.
Now, we have to install PowerPath, which is new to us. The EMC consultant tells me on Monday that the LUN device file names may change when PowerPath is installed. Of course he was not sure after I questioned for a definite answer. I checked the Powerlink (and the ITRC as well) docs and did not find anything that indicated that the existing device file names will change, only that I should remove the alternate pvlink, but these were the installation guides. I do not see a reason for the device file names to change, but I am new to this product. Can I get a definite answer from the Sys Admins here who have installed powerpath on an existing server? For rp7410, rp7420, and n-class servers.

thanks

[Mauro Gatti · Accepted solution]

Hello Etienne,
as far as I know powerpath doesn't change device file.
The main difference is you haven't to configure alternate link (PV link) because powerpath automatically knows which are alternate paths even if they aren't specified in VG configuration.
I deal with L,N,K,D and A class and a CX700 storage system.
Let me know if you find out any other powerpath behaviour.

Regards

[Youlette Etienne_2]

thanks, that is much more reassuring. I was quite amazed that the EMC consultant did not know for sure. From what I've read so far, it provides I/O load balancing as well. Not having to configure the pvlinks anymore does make things a bit easier.

[Mauro Gatti]

check also correct failover policy that would be "policy=CLAROpt" (check it using "powermt display dev=all" command and check it for each clariion LUN)

Bye

[Youlette Etienne_2]

Also, for each initiated host, under teh connectivity status, on the SPs, should the failover mode remain set to "1" after powerpath is installed?

[Mauro Gatti]

This is the most difficult question because EMC technics too about failover mode settings are not so sure :-)
As far as I know if you use powerpath you would have to use failover setup wizard to set all hosts with initietor type "Clariion Open" and failover mode "0"

[Youlette Etienne_2]

Thanks. I think EMC mentioned the same "0" setting, but after this powerpath issue, I was not too confident about that answer either.

[Stuart Abramson]

1. PowerPath doesn't change device file names. The EMC guy should have known that.
2. the Clariion config for volumes with PowerPath is different than for PVLINKS. There are two variables that change:
Intiator Type
Failover Mode

See the EMC Manual "EMC Host Connectivity Guide for HP-UXâ

[Youlette Etienne_2]

thanks. I will check that manual now.

[Youlette Etienne_2]

information was most helpful

[Youlette Etienne_2]

Ok, I finally got PowerPath installed and all devices are active and alive. Now, it takes 45-60 minutes to reboot the server. Below is what I sent to EMC:

PowerPath is presenting another major issue. On system reboot, it takes 45-60 minutes to go through the /etc/lvmtab on the server, i.e, the point in the boot process at which the volume groups are scanned (querying each SAN physical LUN) , reactivated, and the file systems are mounted. This normally takes around 10 seconds, if that much.

Have any of you encountered this problem? I have 32 SAN vg's on this server. Each one takes approximately 2 to 5 minutes to scan, activate and mount. I rebooted the server twice to make sure that this was as a result of some PowerPath 1 time configuration on the first reboot.

[Stuart Abramson]

Ours don't take anywhere near that long to reboot. We don't have as many as 32 volumes on our servers with CX600. Our largest number of LUNs is 8 LUNs of 120 GB each. (We run big LUNs on the Clariion - maybe that's why!) I'll ask the guys that have been here longer.

[Youlette Etienne_2]

I don't think the size of the LUNs makes a difference. These are varying sizes, from 400 Mb to 200 Gb. It takes pretty much the same amount of time for either. Of course, the next task is to get the DBA to stop creating such small LUNs.

I have not come across anything else from in the docs and still waiting for an EMC response.

[Stuart Abramson]

I agree the size of the LUNs isn't important.

It's the NUMBER of LUNS that's important. HP-UX verifies that he has his PVs and quorums at boot and the MORE PVs you have the longer it takes to boot. So to reduce the number of PVs, we make the LUNs bigger. i.e. if you need 1 TB you could have:

1 TB = 1000 GB =
............1000 GB/10 GB per LUN = 100 LUNs
............1000 GB/100 GB per LUN = 10 LUNs

That's why we have big LUNs. Less LUNs per server.

In some cases bigger LUNS reduces performance because you have less LUNs to split your oracle database across, and can do less in parallel. We only use Clariion for "low i/o filesystems". We use Symm for big Oracle databases.

[Youlette Etienne_2]

There is one LUN per vg. Only change, PowerPath was install. From managing a Symm, where I had multiple hypervolumes assigned to vg's, it still took only about 10 seconds to go through this step.

On a side note, from the problem at hand: you brought up an interesting point about the LUN sizes and Oracle database performance. Do you know of any white papers that discusses Oracle OFA best practices or something similar? I have been mainly focusing on the OS/SAN administration and left this up to the DBA, but I would like to know more about this. Thanks

[Mark Fisher_4]

I would be interested in knowing the the slow vg activation issue was resolved? I have been battling this one for months with EMC. The provided a patch for Powerpath, 4.3.1.2 but it did not resolve the issue for me. I rebooted a server last night with 10 luns. Each vg activation took 5 minutes. The more luns, the slower the vg activation per vg. I can't deal with 1 hour reboot times...

[Youlette Etienne_2]

I am highly disappointed in EMCs tech support--i've gotten some truly disappointing and incorrectanswers from them. I ended up finding out what the issue was myself. From the installation guide, after installing PowerPath, the patch that they recommend be installed creates this problem. Just installing the PowerPath alone should resolve the issue. If it does not, you will have to backrev to v4.2.


Another Question for All:
Has anyone experienced a siginificant I/O performance improvements with PowerPath? Were there any CPU cycles trade offs? From a brief look using our performance tool, I/O was balanced across some channels (about 3 or 4), but not all 8 per LUN as I was expecting.

[Mark Fisher_4]

So, running Powerpath 4.2 resolved your issue with the vg activation times? If so, I'll have to find a copy and give it a try. So far, I've tied 4.3, 4.3.1 and 4.3.2.

We originally had load balancing problems but the SAN folks were able to setup the default paths via Navisphere. The only issue I have seen is if they have to reboot SP A to install a patch, then all the luns failover to SP B and stay there. Then your have to rebalance the luns with powermt restore.

[Vic S. Kelan]

Hi All,
I know this is a very old thread but just to say I upgraded to powerpath 4.5 from 3.0.3 AND my device file names CHANGED!

I read this post before the upgrade and would just want others to know it could change.

[Youlette Etienne_2]

Thanks for the heads up Vic. I attempted to install v4.5 on our rp8620 IA 11iv2 server, and it showed all of the connections as "alive:dead". As a result, I could not communicate with the LUNs. I uninstalled it. I still haven't gotten v4.3 to work on the n-class servers, only on the rp74x servers.

[Adam Mullen]

We're upgrading from a Symmetrix DMX800 to a Clariion Cx700. One response I got from the "san engineer" was that 'You might need to upgrade powerpath...' When I asked him why, he wasn't able to provide an answer. I'm currently at 4.2 and frankly, after the reading I've done, plan on staying there. Just my two cents. Yes, I'm disappointed with EMC too!