I believe to be about to squash this bug from 2002: http://sourceforge.net/tracker/index.php?func=detail&aid=563486&group_id=12694&atid=112694
I described it more concisely in the subject to this mail than the subject of the bug says - I hope. From the comment I added to the bug: --- I have a snmp daemon that works as a proxy for another snmp daemon over *TCP*, ie it forwards requests. In case the second snmp daemon goes down (or one of the daemons is on a DSL line and gets a new ip address), the TCP connection breaks. The first snmp daemon never notices that and a query with snmpget will exit with the error "Timeout: No Response from xxx.". To fix this, the first snmp daemon should try to re-establish the connection. I guess the code in snmplib/snmpTCPDomain.c should be extended to perform such reconnect. --- Please advise me whether my plan is reasonable: 1. Check the return values of send() and recv() for timeout or other conditions to find out when the TCP connection has gone dead (How to see that?). 2. Try to re-connect (Any strategy for that - ie. limit the number of tries per minute?). Please forgive for crossposting the comment and this email, but net-snmp-coders has much more subscribers than net-snmp-bugs. Kindest regards, Michael Bunk
pgpgRq07WYm2J.pgp
Description: PGP signature
