ADSM-L

Re: ADSM 3.1.2.40 DB unload/reload

2000-09-26 08:59:35
Subject: Re: ADSM 3.1.2.40 DB unload/reload
From: Reinhard Mersch <mersch AT UNI-MUENSTER DOT DE>
Date: Tue, 26 Sep 2000 14:58:47 +0200
Werner,

I _DO_ think, that insufficient space actually is the reason. I had a
similar situation last year, when I tried to load a database originally
sized less than 25 GB (i don't remeber the exact numbers). I needed
several attempts and only after spending more than 40 GB, the LOADDB
succeeded.

The reason, as I was told, is: LOADDB uses the traditional B-tree page
split algorithm, which results in ~ 50% filled pages on the average, while
the server software uses a modified split algorithm resulting in a higher
page utilization. So, when loading a DB, you might need twice as much
space. I complained loudly about that "feature" last year. It would be
nice to hear, whether something has changed in 3.7 or 4.1.

Good luck, and tell us about your experiences.

Reinhard

Werner Baur writes:
 > For sake of shortness I have described our problem not very exactly. I 
 > should have
 > mentioned that we have first tried db backup/restore. Unfortunately restore 
 > db failed
 > several times. We have been told by IBM support that the reason for the 
 > failure is an
 > inconsistency in the database and that audit db is required.
 > Nobody can tell us how long an audit would last and our experiences with 
 > audit db in
 > the past are not very promising ...
 > Right now the load db failed after 4 days with an "insufficient space 
 > message".
 > Actually this was our third attempt to load the db. The loads before ended 
 > up also
 > with "insufficient space messages. I doubt that insufficient space is the 
 > reason for
 > the failure, we have provided 90 GB for loading. The original size of the db 
 > is 69 GB.
 > In fact I suppose that the assumed inconsistency of the db causes the 
 > problems.
 >
 > Originally we provided backup/restore db as fallback in case of problems 
 > with the
 > export volume group approach. Now it looks like we have to forgo the 
 > fallback. At
 > least it is encouraging that others have done a successfull server movement 
 > before.
 >
 > Thank you all for your hints,
 > Werner

--
Reinhard Mersch                        Westfaelische Wilhelms-Universitaet
Reinhard Mersch                        Westfaelische Wilhelms-Universitaet
Zentrum fuer Informationsverarbeitung - ehemals Universitaetsrechenzentrum
Roentgenstrasse 9-13, D-48149 Muenster, Germany      Tel: +49(251)83-31583
E-Mail: mersch AT uni-muenster DOT de                       Fax: 
+49(251)83-31653
<Prev in Thread] Current Thread [Next in Thread>