multiple instances of "ctl_cyrusdb -c" hanging lmtp (?)

[Chris Stromsoe]

I occasionally get multiple instances of ctl_cyrusdb running.  When that
happens, lmtp connections hang, resuling in processes that don't do
anything.  Doing a stop/start is the only way I've discovered to get
things moving.

Am I correct in assuming that more than one "ctl_cyrusdb -c" either locks
deliver.db for excessive periods of timing causing extreme slowness or
that it deadlocks, preventing lmtpd accesses?  If this is the case, would
there be a problem implementing an addition flag in master.c:add_event()
to limit the maximum children an event can have?

cyrus 2.1.9 under debian linux system, with BDB 3.3.11.  deliver.db is
db3_nosync, mailboxes.db is skiplist, seen is flat file.  lmtp delivery is
via a TCP socket.



-Chris