1.6.0a3 Denying repeated connection and Copy type

[Jerry Christopher]

I have been searching gnu.cfengine.help for someone with a similar
problem, but now I turn to the list...

Using cfengine 1.6.0a3 on the server and client, Solaris 2.7 on both.

We were getting the following in /var/adm/messages on the server:

Mar 11 07:45:32 gold gold[11705]: Denying repeated connection from
10.101.101.25
Mar 11 11:45:31 gold gold[11705]: Denying repeated connection from
10.101.101.25
Mar 11 14:45:29 gold gold[11705]: Denying repeated connection from
10.101.101.25

The log messages coincide with 8:45, 12:45, and 15:45 cron entries on
the client which resides one timezone away.  The cfengine.log on the
client shows the denied connection from the server.  So these runs go
across the WAN.  We determined the link does not have the bandwidth for
all the updates that were going across.  Several hundred Mb were being
transferred and the current run was not finishing before the next
started.

Further investigation showed that files that were up to date on the
client were being transferred each day.  I'd look at our cfengine.log
and see that the 8:45 client run "updated images" for files that
definitely haven't been changed on the server in months.  I commented
the cron entries and let this run to completion - sometime after 12:45,
but before 15:45.  After the 15:45 run, the log shows the files not
being updated as we would expect.   In the morning, 8:45, the "updating
images" for the exact same files happens again and several hundred Mb
are started for copying across the WAN.

>From the cfengine reference, I read that cfengine uses the ctime
date-stamps on files to determine whether a file needs to be copied: a
file is only copied if the master is newer than the copy of if the copy
doesn't exist. Here is one copy statement intended to copy the trees
beneath /distrib/desktop to / on each client. 

        /distrib/desktop
                dest=/
                server=$(MYDISTRIB)
                backup=false
                recurse=inf


>From what I've been reading in the newsgroup, should we be using
type=checksum or byte rather than the default of ctime?  

We want the client to get any file changed on the master, plus the
client should be updated if it's copy differs from the master.  Sounds
like checksum is what we are after.  If you are using type=checksum or
byte, can you please comment briefly on performance over the default of
ctime?


Thank you for your time,

Jerry Christopher
Applied Micro Circuits Corp.
6290 Sequence Dr
San Diego, CA 92121