SmartArray 5304 adding hot spare

[Graham Allan]

We have an ES40 with SmartArray 5304 for its system drive (4x 72GB drives, RAID6).

I was foolish enough to set this up without a hot spare; now one of the drives has been marked as failed: not a big problem; I swapped in a new drive. But I have two issues:

1) It appears there is no way to rebuild the raid set from within Tru64. It does offer to rebuild when running the RAID BIOS during console init, but any further init seems to cancel the rebuild.

1b) (connected to the above) Maybe I am missing some SRM setting; it seems that if I have AUTO_ACTION=HALT, then when the system completes its very long self-test and I enter "boot" at the prompt, it performs another init and slow self-test before booting. Is there any way to disable this? It seems very unnecessary!

2) I can't see any way to define a drive as hotspare - is it really not possible to add a hotspare to an existing array?

The SRM version is 7.3-1 and Tru64 is 5.1B/PK6

Thanks for any advice,

Graham

[marsh_1]

hi,

adding a hot spare to an existing array is supported , which version of acu are you using to try and configure the raid card ?

hth

[marsh_1]

hi,

if you don't have acu all the downloads and docs for it can be found here :-

http://www.compaq.com/alphaserver/products/storage/sa5300a/index.html

hth

[Graham Allan]

Thanks, I did have ACU installed (package CPQACUXE13071a) but I obviously (now!) failed to fully configure it, as I could view array status in the management web pages but not access any administrative functions.

Now that I reread the install docs and create an IM_admin group, I do get an "Other agents: ACU-XE" link on the management page. Unfortunately it doesn't work, as when I click it I get a new window which eventually gives me a frame set of multiple "Not found" messages (eg "The URL /ACE-XE/ACU-Body.htm was not found").

I guess there must be something not configured correctly within the management pages to link its web server into the location of the ACU content.

The version of the management page/agents I have installed is CPQIM370, which seems like the latest version... is there any kind of incompatibility or install order requirement?

[Graham Allan]

Hmm, having played some more with the ACU-XE web agent, I don't believe it is a config problem. It gives me different results each times; reloading the page over and over gives slightly more each time until eventually the whole thing works. Possibly it's using some dodgy javascript, though I started out with firefox 3.5, then IE7, then went all the way back to firefox 2.x on the server itself, none of which really worked smoothly.

But, unfortunately even once it works, the ACU-XE agent on tru64 seems to have no very useful options. You can set a few overall controller parameters, and as for the arrays themselves they can be created or deleted, but nothing useful can be done with them.

This is quite disappointing as clearly the controller has the ability to do much more. the documentation shows that the CLI (apparently only available for windows/linux) does let you assign hot spares, or migrate arrays between raid levels, and many other things.

I really hope I'm missing something and there is some kind of CLI utility somewhere for Tru64...!

[John Manger · Accepted solution]

An answer to '1b' is to set BOOT_RESET to OFF. That prevents the 'init' when booting. However, it is not usually recommended to set it to OFF in a cluster.

John M

[Graham Allan]

John, thanks for the tip about BOOT_RESET. This system isn't in a cluster so I guess it shouldn't cause problems. Our main ES40 has 16GB memory so the self-test on init takes an eternity. The wait isn't so bad on the 4GB machines!

Graham

[Rob Leadbeater]

Hi Graham,

The other answer to 1b is to change the memory_test SRM variable to be OFF or PARTIAL.

http://h18002.www1.hp.com/alphaserver/download/es40fg_revb.pdf (page 94 / 2-76)

P00>>> set memory_test partial

Cheers,

Rob

[Graham Allan]

To summarize, it does seem that the ACU-XE utility on tru64 is... hmm, not very useful.

But I am very happy to have eliminated the slow self test during init. I opted to use the "BOOT_RESET=OFF" setting rather than "MEMORY_TEST=PARTIAL", as some other postings in the forums referred to potential weird behavior on the ES40 if this setting is changed.

Thanks,

Graham