cdcc info and rtt don't query all servers in the map

Vincent Schonau
Fri Oct 29 15:28:40 UTC 2004

On Oct 29, 2004, at 16:28, Vernon Schryver wrote:

>> #      25% of  8 requests ok   10.49 ms RTT              8 ms queue 
>> wait
>> Note that cdcc claims that was 'not answering' at
>> this time. It most definitely was. A tcpdump of traffic to or from
>> from the host that this command was run on, shows no
>> packets being sent to that server at all. A random check indicates 
>> that
>> more of our dcc client installations exhibit this behaviour, although
>> apparently not all.
>> Is that intended behaviour? Does this not cause that last server to
>> never be picked by the client as the server to send it's reports to?
> How do you figre that the fifth system was in fact answering?

Because it looks fine from other hosts, which have dcc1 through dcc5 in 
a different order in the map.
It's not specific to this host; it's specific to the 5th entry in many 
maps in our client cluster.

> Did `cdcc "host; rtt"` get answers?

I had not tried that. That does get an answer:

# cdcc 'host; rtt',-        anon
# *,-                                           XS4ALL 
ID 1118
#     100% of  1 requests ok   50.38 ms RTT             50 ms queue wait

> Are you running FreeBSD 4.10?

Sorry, I should have specified:

$ dccproc -V
  $ uname -rmps
FreeBSD 4.10-RELEASE i386 i386

> This week I finally got around to doing
> somet tests with FreeBSD 4.10 and found that UDP socket disconnecting
> is indeed newly broken in FreeBSD 4.*.  One symptom is that `cdcc rtt`
> overlooks all servers after the first one that does not answer.
> I will release 1.2.58 with a work-around later today.

Excellent; I will try that, then.

Vincent Schonau

More information about the DCC mailing list

Contact by mail or use the form.