HSG80 Config Save Problem

[John Kinnaird]

I am having a proble trying to save the configuration on a Single Controller System. I have initialised a Drive with the save_config flag, This shows as ok when I type show dev full ( or show disk20400) I have set nofailover when I type config save I get Error 2112: Configuration Save not enabled. I have tried a config reset and restart without making any difference.

[Mike Naime]

There was a VCS patch that disabled the SAVE_CONFIG switch on the 8.6F-? firmware. ITwas causing controller crashes, so they disabled that functionality. We had a similar issue since we used to alway initialize at least one storageset on each HSG with the SAVE_CONFIG flag.

[Uwe Zessin]

Mike,
a VCS patch for ACS ? ;-) Yes, yes, I get confused sometimes, too :-(

But you know that SAVE_CONFIGURATION is only supported on single-controller configurations, do you? A field service person from Compaq once told me that it was only supported on a JBOD, but I have found nothing in the manual.

Anyway, shouldn't it work on a newly initialized storageset? Just an idea: would it make sense to set the disk to TRANSPORTABLE and run DILX over the whole range of blocks?

By the way, I have noticed that the patches -2 and -3 for ACS V8.7 have been reloaded between october (when I checked them first) and today. There is now a .PDF file included.

[John Kinnaird]

This is in a Test System, It has only 1 Controller running at ACS v8.6-8 We have a Live box which has a controller at Hardware level E06 and should be E08 min. I cannot run Dual Redundant as I've only 1 ACS Program Card a cache module. My plan was to save the existing config. Install the new controller and restore the config, Then soak test it for about a Week prior to installing it in the 'Live' System. Easy Huh ! I cannot find any info on Error 2112. I've also tried it at ACS v8.6-13 with no difference.

[Uwe Zessin]

John,
I have no idea how to decode that 'security code' from 'SHOW THIS_CONTROLLER' and I have not found any more information about 'Error 2112', either.

However, I notice that you have 3(!) disks with configuration backups enabled. It might be that this is the problem.

Can you just delete the units, storage sets and disks from the controller configuration and retry with only your single disk?

[John Kinnaird]

I have tried deleting all units reinitialising every drive as a JBOD ( No Save_config ) adding all drives to Units as JBODS, Then I ran DILX on all Disks for 1 Hour - No Errors I then reinitialised 1 Drive ( Disk10000 this time ) with save_config and guess what - No Change I'm beginning to think I may have a Hardware Fault, But I'd like to find out more info on the 2112 Error !

[Uwe Zessin · Accepted solution]

I see that you ran these tests with ACS V8.6-8.
Honestly, I am too lazy to check which patch disabled the config support, but would you mind to retry with patch -13?

[John Kinnaird]

Worked a treat ! One last question, The Live box currently has 2 Dual Controllers at ACS v8.6-8 What if any are the implications of running with one at -8 and one at -13. ps I think the fix is patch 11 ! Thanks again - John

[Uwe Zessin]

Hey! Great! Finally...

John,
I would not run controllers with different patch levels, especially not _that_ much different. There could be differences in the metadata. If the newer controller disables the configuration data and the other is not aware of this. Well, you should not SAVE_CONFIG on a dual-controller system anyway, but there could be other areas affected.

What is the reason that you want different levels?

[Mike Naime]

With dual controllers, there is nothing to prevent you from saving the configuration with one controller, and then restoring from the other controller. Because of this fact, the save_configuration on a dual controller configuration is not supported. Compaq did not want to be responsible for all the users out there that would save the configuration using one controller, and then restore from the other controller. (Gee, Why did I loose my data? You did not prevent me from loosing my data... Therefor you must be at fault. It can't be that I do not know what I am doing.)

If you know what you are doing and have some standard like always saving the configuration from the TOP controller and NEVER restoring from the bottom controller, then you are Ok!

[Mike Naime]

Another Comment. If you initialize a raidset or mirrorset with the SAVE switch, it will keep the data on all members of the storageset.

Having multiple coppies is not a problem. It is redundancy to ensure that you have the data when you need it.

[John Kinnaird]

Just to Clarify my thinking, We have a Live and fairly critical system running v8.6-8 acs on both controllers. I have had a couple of instances in the past where Compaq / HP replacement controllers Failing after a few Days. My I also have a single Controller 'Testbed' System. One of the controllers in the Live System is running at Hardware Rev E06 and should be E08 or above . My thinking was to save the configuration on my testbed, Install the replacement controller Restore the config and test it for a Week or so prior to fitting it in the Live Box. My only reason for not upgrading both controllers to 8.6-13 is that should the replacement controller have any problems, The Origional untouched and trusted controller will take over, However after reading your comments I'll try and get approval to upgrade both controllers to 8.6-13. I will odf course use set failover copy=this on the Dual redundant system as config save is only supported on single controller systems. Sorry for any confusion. Thanks again for all your Help. ps I'm off skiing next Week so if you require any more info, I'll ket you know when I get Back !

[Uwe Zessin]

Mike,
sorry, but I don't understand you comment about 'save on one controller and don't restore on the other'. Doesn't an unconfigured controller always go out to the disks and try to grab a config?

I had tried such a situation where I put an unconfigured controller in an M2200 enclosure that was powered down. After power-up, this controller loaded the config from one of the disks and cause an invalid cache status. Oops! Where is my data ;-)


Controller replacement on HSG systems is really a pain. In case somebody is not aware: there are release notes available for some ACS V8.7-x versions (AA-RUQ2A-TE). They are now included in the -2 and -3 patch files. See:

http://h18000.www1.hp.com/products/storageworks/softwaredrivers/acs/index.html


John,
I would not try to transfer configuration data from one storage system to the other. The controller WWN will move, too, however the metadata on the disks is wrong.

You say you have a fairly critical system, shouldn't you better use MULTIBUS failover rather than TRANSPARENT or was that just an example you wrote?

Have fun skiing and come back in one piece!

[Mike Naime]

Uwe:

There are two CLI ports on the front of the HSG. One in each controller. For my environments, we refer to these as "TOP" and "BOTTOM" even though the EMA cabinet has 3 of these on their sides. :-) If you do all of your CLI commands through the same port (We exclusivly used the TOP port for about 2 years) then you have no concern about where the configuration save came from.

During that timeframe, I fried an HSG! We lost an entire backplane, both HSG80 controllers, 2 EMU's and the Cache modules. Luckly this was an 8-5 client running on that controller, and we had overnight to restore everything.

When we replaced the BA370 backplane, plugged in new controlers... Etc. We were able to restore the configuration from disk by using the TOP controller first. After that controller came back online and was functional with the correct configuration. We then added the second controller. (Set Multibus copy = this) We got our entire configuration back, and did not have to re-key anything. All of the data was still on disk as well.

You will get the Invalid Cache message like you do when replacing most failed controllers. After you clear that message (Clear this invalid_cache no_destroy.... Etc) then you can proceed. Note: We had a Compaq FE onsite assisting us with the restore.


I have an entire HSG that I reclaimed last week. I can try initializing some of the drives with the SAVE_CONFIG and see what happens.

[Uwe Zessin]

The document whose pointer I posted above claims to describe a way how to prevent the cache invalid messages during replacement.


I will never forget how to clear this condition! It was in late 1996 when I was tasked to set up a HSZ-20 with two AlpaServers running Windows NT and 'Digital Clusters for Windows NT'. Somebody had forgotten to shutdown the array and turn off the cache battery before packing.

I had not done much with the HSZ up to that date, although I had worked with HSC50, HSD05 and HSD10. Well, the documentation was pretty helpful and I managed to get out of that situation myself - thank you, Digital!