Strange message from cron-dccd

Vernon Schryver
Mon Oct 7 00:50:34 UTC 2002

> From: Sean Rima <>

> ...
> Oct  6 14:34:06 bsod dccd[248]: database cleaning begun
> Oct  6 14:35:57 bsod dbclean[5878]: expired 117719 records and 208989
>                 checksums in /var/dcc/dcc_db
> Oct  6 14:44:58 bsod dccd[248]: "ADMN FLOD received before ADMN
>                 DB_NEW"  sent to client 1042 at,1236

That is a serious error that says one of the messags from dbclean to
dccd was lost.  Since both programs must be running on the same
system, that implies that something is seriously wrong with the
loopback network on your computer.

That implication of something hosed with lo0 routing is consistent
with other problems you have reported in the past.  

Is there any chance that sometime between 14:35 and 14:44 your PPP
link went down and then came back?

That 11 minute delay is not good unless you have a small system
with slow disks.

> Oct  6 17:08:24 bsod dbclean[5878]: bad hash free list entry at 0x2
> Oct  6 17:08:24 bsod dbclean[5878]: failed to relink record at
>                 0x314c7cc;  fatal error
> Oct  6 17:08:27 bsod dccd[248]: 1.1.14 database /var/dcc/dcc_db reopened

Those errors are bad, but no worse than you'd expect with the
previously logged problem.

> ...
> I remove the existing database and created a new one, since then dccd
> has run okay. Something may have happened when I copied the files up
> the new location, I tar.gz'd them. ...

What files were copied?  I thought you said that you copied only the 
DCC source and then built it.  

Copying the /var/dcc/dcc_db file is unsafe if dccd is running when
the snapshot is taken or if the source and target computers differ
significantly, such copying from a SPARC to an 80*86.

There's no good reason to copy the /var/dcc/dcc_db.hash file instead
of letting dccd automatically ask dbclean to rebuild it.

Vernon Schryver

More information about the DCC mailing list

Contact by mail or use the form.