ADSM-L

Re: mirrored DB

1996-05-29 10:25:22
Subject: Re: mirrored DB
From: Mike Wilson <mwilson6 AT TUELECTRIC DOT COM>
Date: Wed, 29 May 1996 09:25:22 -0500
Andy,

We stopped mirroring our database after we moved all ADSM files to RAMAC II
DASD. Would you agree with this?

Mike Wilson
Storage Administration
Texas Utilities
At 05:47 AM 5/29/96 PDT, you wrote:
>John O'Neall asks:
>
>>Can someone tell me in what case(s) the mirrored copy of the DB (and log)
>>is supposed to help.
>>
>>I tried an experiment, since we're not yet in production.  I stopped
>>ADSM, then unmounted the file system on which one of the DB mirrors
>>resides.  Relaunching ADSM got it as far as the message
>>
>>ANR7850I Process 39958 has terminated on signal 13 (Write Pipe, No Reader).
>>
>>at which point it hung without starting any more than 2 dsmserv
>>processes.  It was impossible to connect an adminsitrative client and
>>therefore to do anything else.
>
>John,
>
>I'm not sure why we aren't recovering from this error. Is it possible that
>the dismounted filesystem might contain other files that either ADSM or the
>operating system needs? If you can't find anything like that, you should open
>a problem with IBM support.
>
>In the mean time, try re-mounting the filesystem, then renaming the mirrored
>DB copy and see if ADSM can get by that.
>
>By the way, although the ADSM product supports the use of raw logical volumes
>for database and recovery log files, we generally don't recommend their use,
>as the benefits do not usually outweigh the potential problems. I don't know
>whether you are using them or not, but I just thought I'd mention this.
>
>As to what benefit mirroring provides: mirroring helps prevent loss of the
>database or recovery log in the event of a media failure. This doesn't
>necessarily mean that ADSM might not abend in the event of the failure, or
>have problems starting up. You may have to take some kind of corrective action
>in order to get ADSM started up again, depending on the nature of the problem.
>For example, if there is a bad page in the primary copy of the db or log, you
>might have to set MIRRORREAD LOG and/or MIRRORREAD DB to VERIFY, which will
>force ADSM to look in the mirror copies for the good page, at which time all
>mirrored volumes will be resynchronized. (By the way, in this example, once
>ADSM is successfully restarted, you should shut it down again and set the
>MIRRORREAD LOG and MIRRORREAD DB values to their original settings, as VERIFY
>usually involves a performance hit. Then restart ADSM.) The point is, you've
>saved yourself from the very painful dumpdb/install/loaddb/auditdb process.
>
>Andy Raibeck
>ADSM Level 2 Support
>408-256-0130
>
>
<Prev in Thread] Current Thread [Next in Thread>