[Imap-uw] hung imapd daemons
ketcham at chem.washington.edu
Thu May 6 20:29:55 PDT 2010
I run IMAP4rev1 2007a.403 on RHEL 5, from xinetd. I find imapd
processes running under uid root that hang around indefinitely and
gradually accumulate on the mail server. netstat shows that these are
connected to known client addresses, and this is verified by log records
showing the initial connections that spawned these processes from known,
I have seen on a few occassions that these root-owned daemons accumulate
to the point of maxing xinetd limits, either the total processes limit,
or the per_source limit. In the case I recently investigated most (but
not all) of these daemons were started up from one particular client
address, and the user at that address was reporting a thunderbird
message about "too many connections". Ten root-owned imapd processes
were found to be connected to her comcast ip address, (ten being or
xinetd per_source limit) and the log records showed some activity in the
middle of the previous night from that thunderbird client, first some
autologout messages, and then a series of connections every 10 minutes.
I asked the user and she told me that she leaves her computer on at
night, but she was not working on it at that time. I don't know of
anything unusual in that user's setup.
As far as I understand this problem, the client initiates an SSL
connection, but for some reason the initialization does not complete.
The newly spawned daemon hangs before it changes uid, and just hangs
around indefinitely. I can locate records in our mail log for these
connections and I see the message: "imaps SSL service init from
xxx.xxx.xxx.xxx", but not the authentication message which normally follows.
Can anybody offer an explanation?
U.W. Dept. of Chemstry
More information about the Imap-uw