Re: Timeout problems, also more detail and thoughts

[coffeelover]

Olivier and others,

> I have installed sendmail 8.12.11 with the tarball 

 we are not using the same platforms so what worked for
me may or not help you. ??
 I did not use a tarball to install from. I built custom
rpm based binaries. I actually used the latest source
rpm for redhat 8.0 fetched from updates.redhat.com
 I used this method to build binary RPMs for both
redhat 6.2 and redhat 7.3 ... using each platform to 
generate the binaries for each. ie: i used an up to date
redhat 6.2 to build custom binary sendmail RPMs and of
course i used redhat 7.3 to generate the same for my 
redhat 7.3 servers. 
 I then ran into the milter problems only on the redhat
7.3 servers. the five redhat 6.2 servers did fine. No
need to mention everything i tried. I think i tried to
find every possible combination that did not work before
i noticed that the redhat 6.x patch did not define IPv6
while the patch for 7.x did along with a couple of other
extra defines. Also for anyone following along be sure
to specify 7.2 and not 7.3 in the sendmail.spec file
and make sure that milter is defined somewhere either
in the patch file redhat uses or customize the spec
file. Bottom line is removing the IPv6 define from the
patch file applied to redhat 7.x versions solved ALL !!
my problems.
 some points:
 1. i used redhat's rpm system to install an updated
sendmail for both redhat 7.3 and 6.2 servers. this 
method is actually far easier than building from tar
balls or so i think. you do need to learn how the rpm
build "stuff" works for 7.x and earlier you use rpm -bb
for 8.x and later i think? you use rpmbuild or is it
buildrpm command ??? Not running any production 8 or 9
servers since my focus and needs are stability
 
 2. redhat 9.0 along with enterprise 3.0 has the linux
native threads stuff patched/hacked in. I wonder if
this could be causing problems. also applies to fedora.
when i run spamass-milter i see three or so of them
startup ... kind of threading i understand. The milter
would hang if they were not "all" there. I did a simple
watchdog script that looked for at least three spam-ass
milters and if not then kill any remaining ones off and
restart them. The restart script would be triggered
several times per day on a rather not busy server ...
much less than 1000 messages per day. the redhat 6.2
boxes have yet to have the milter restarted. After i
rebuilt the 7.3 RPMs without ipv6 defined the 7.3 boxes
are now as stable.
 Note: i have not run spamass-milter in any kind of 
redhat 9.0 environment. i intend to stay with my very
stable redhat 7.3 boxes for the remainder of this year
and probably until summer of 2005 before considering
any changes.
 hope this may help you or anyone else a little bit
-Patrick Main-


>   Compiled with: DNSMAP LOG MATCHGECOS MILTER
> MIME7TO8 MIME8TO7 NAMED_BIND NETINET NETUNIX NEWDB 
> PIPELINING SCANF USERDB XDEBUG