database broken

> From: Ken Serrine <kserrine@cat.com>

> I'm getting a "database broken;" at least once per day now.  My dcc_db 
> is over 2 GB and has been for a while.
> The automatic repair runs, but it really bogs down the box in the middle 
> of the day.
> Is this typical of anybody else?
>
> I run dbclean every night and I've tried setting
> DBCLEAN_ARGS="-E 7 DAYS"
> in the dcc_conf, but it doesn't have any effect.

I bet that if you find the various messages that dbclean always
sends to the system log, they will include a complaint like

    Jan 12 16:44:10 calcite dbclean[57412]: invalid long term spam expiration seconds "-E 7"; fatal error

The same message should also be mailed by the cron daemon to the mailbox
associated with whichever user runs /var/dcc/libexec/cron-dccd

Because that is a fatal error, dbclean is not being run by the cron job
at night.  Instead it is probably run by the emergency work-around-
missing-cron- job-around machinery in dccd.

To stop the error message, try
DBCLEAN_ARGS="-E 7DAYS"

To shrink the database, try something like
DBCLEAN_ARGS="-E 7DAYS -e12HOURS"

However, instead of doing that I would switch to 
DBCLEAN_ARGS=
and install the current version by running /var/dcc/libexec/updatedcc. 
The three DCC servers I know about at cat.com are running version 1.2.50.
There have been many changes since 1.2.50 including many related to
limiting the size of the database.

Note that explicit DBCLEAN_ARGS=-E... settings disable some of the
most important automatic database size limiting mechanisms.


Vernon Schryver    vjs@rhyolite.com