Vernon Schryver
vjs@calcite.rhyolite.com
Sun, 22 Jan 2006 20:55:27 -0700 (MST)
> From: PeakPeak Customer Service > There have been a couple of unexpected exits of dcc. I need a script that > restarts it. > > I use something like this on BIND but it doesn't seem to work too well for > dcc, any ideas? Dccd should *NEVER* suffer "unexpected exits" except when things are too hosed to continue such as with an overflowing disk. Dccd tries to keep running when things get somewhat sick including trying to run dbclean. I think that when dccd quits, an automatic restart is likely to make matters worse. I think it is better to ring alarms so that someone can look at the problem and fall back to other DCC servers, such as the public DCC servers. That is why dccd does not have the restart-after-bad-signal kudge that dccm and dccifd have. However, juging from the use of `daemontools` and so forth, other people may disagree. > I use something like this on BIND Have you considered running a more recent version of BIND? BIND 8 and 9 run for years for me without crashing. > but it doesn't seem to work too well for > dcc, any ideas? > # /var/dcc/libexec/rcDCC start > /var/dcc/libexec/dccd -i1350 -n someplace.com -u100 > /bin/logger -p emerg "dcc restart" > mail -s 'dcc restarted!!' some pager@someplace.com < /dev/null > sleep 62 There are a whole raft of reasons to *NOT* do that. If you insist on automatically restarting dccd, look at the /var/dcc/libexec/start-dccd script to see how to determine whether dccd is still running. Then instead of restarting it ad hoc, use start-dccd. Vernon Schryver vjs@rhyolite.com