Exchanging dccd ids & passwords

Vernon Schryver vjs@calcite.rhyolite.com
Tue Sep 4 17:18:35 UTC 2001


> From: "Mediratta, Bharat" <bharat@fusionone.com>

> ...
>   90109 hash entries  64187 used   3307512 DB bytes

> www.menalto.com gw-menalto 198.144.206.35,6277
>         server-ID 1003  /usr/local/dcc/map  09:47:06
>     version 1.0.27  DB locked  tracing ANON CLNT 
>   40957 hash entries  29815 used   1935960 DB bytes

> >From the "hashed entries" lines it seems like I have a lot 
> less data in my map than is on calcite.  Why is that?  Will my
> server be less effective than calcite because of it?  Thanks.

The difference is accounted by messages reported to my server with
total fuz1 or body checksum counts smaller than the dccd -t thresholds

It occurred to me that the default value of 20 is too high for my
system, so I restarted my server and caused the flood to your system
to restart from the beginning.  That modestly increased the number of 
checksums you have to:

  40957 hash entries  33626 used   2165412 DB bytes

Note that the goal of flooding checksums among servers is not
to give all servers all checksums, but to quickly distributed
the checksums of bulk mail.  The difference between your 33,626
checksums and my 64,187 can be explained by assuming that about
4000 messages that are not bulk have been report to my system in
the last week, the default `dbclean` expiry for non-spam.

Each message has 8 checksums, so 4000 messages would require about 
32,000 hash table entries.

I think I should change the default dbclean spam expiration threshold
to the same value as the dccd flood threshold.


vjs



More information about the DCC mailing list

Contact vjs@rhyolite.com by mail or use the form.