Subject: | |
From: | |
Reply To: | |
Date: | Wed, 25 Aug 1999 10:32:23 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Re:
> We use Netbase which communicates between its various processes via a series of message files. Occasionally, and not just after a system abort, our main production system was getting major impede problems, traceable to the main Export process from Netbase. Using glance we traced the impede to a "DISK FILL" wait on the first message file - and when we rebuilt the file this problem disappeared - so the problem was not Netbase, but MPE.
>
> The moral - there are clearly ways in which message files can get corrupt - I don't know if MPE engineers have low-level diagnostic tools to look at these files. We take the safe approach of rebuilding our message files as frequently as possible.
The second moral: hit <return> in your mailer, or configure it to
fold lines at 72 columns or so :) (Your first paragraph was one long
line of about 420 characters)
Anyway...the problem isn't merely MPE. Yes, there's a performance
problem with FCONTROL 6 on message files. But, you have to ask:
why is Netbase *doing* FCONTROL 6 on them, and why is the message
file so large? (the problem scales with the size of the message file)
It turns out there's a lot of discussion right there, but the bottom line is:
1) purge/rebuild your netbase message file (sorry, forgot it's
name) at system startup;
2) monitor its size. If it gets large, then netbase isn't keeping
up with your transactions for some reason ... contact Quest and
investigate & fix this.
3) use the current version of netbase, which automatically purges/rebuilds
the message file at startup (sorry, forgot the version)
4) look into testing the new version of netbase, which doesn't use
message files (ditto).
--
Stan Sieler [log in to unmask]
http://www.allegro.com/sieler/
|
|
|