Problems with Relational index (AREV Specific)
At 29 DEC 2004 07:33:59PM David Craig wrote:
Arev knowledgable;
We have a problem with one of our relational indexes. When whatever workstation is indexing hits a certain record it will get the message: "The row RTP57 in file 0032000000000093CLIOTMP\REV69095.LK00047DICT.MFS is over 64K bytes limit for updating relational indexes. Relational indexes for this row will not be updated." then, of course any other workstation that tries to index gets the "another workstation is currently updateing indexes on table 'articles' with control column 'entry_number_xref'…" message and locks up that workstation as well.
What I've done so far: searched the indexed field in the detail table for long values, there were none over 9 characters. Then I removed and rebuilt the indexes - this almost always fixes our indexing problems, but towards the end of a 2 1/2 hour rebuild the same error message popped up repeatedly.
Any help is welcome at this point.
Thanks in advance,
David Craig.
ABC-CLIO
At 29 DEC 2004 11:26PM Curt Putnam wrote:
In whatever table the indexed values are stored, there is a record that is exceeding the max length of appx 65000 bytes. Just list the master table record lengths and you should find the culprit.
Identifying the problem is simple; fixing it is another kettle of fish. The obvious answer is that you should convert those relationals to btrees. That could be painful. As a potential "quick fix" consider splitting the master record such that the details will index into one or more master records - you could do this be indexing a symbolic that checks the master record length and then adds a *1, *2, etc. as required to index to a "new" master record.
If only one record is overlong, it could be a data problem - worth a check.
At 30 DEC 2004 02:57AM [email protected] wrote:
In addition to what Curt says, if this does not resolve the problem we recently came across something very similar due to an unusual corruption so feel free to drop into email. It is likely that this is not the cause as this was with a bug in an old 2.1 system but…
The Sprezzatura Group Web Site
World Leaders in all things RevSoft
At 30 DEC 2004 02:34PM David Craig wrote:
Thanks to you both. I did manage to find the record but after some research found out that that isn't being used in current processing so I just removed it. I think it was used when bringing the system up or in some early project.
Anyway, problem solved for now, thanks again.
David C.
At 31 DEC 2004 09:34AM Warren Auyong wrote:
There's also a glitch in the index rebuild process for relational indexes in 3.12. It will attempt to build an index on null values thus the rebuild can blow up with max variable length exceeded error even when none of the other records hit the 64K ceiling.
At 03 JAN 2005 08:32PM David Craig wrote:
Thanks, that may explain why the symbolic indexed generates 5 0's if the field is blank. We recently had some changes to our processing that results in a lot of records with nothing in that field and after digging around I figured out that the record that blew the limit was an index containing all the ids of records with 5 0's in the field.
I don't like Arev problems but I always learn something, thanks again.
David C.