Hello, I know it is almost christmans, everyone probably are way out to holidays...
But. I have problem with dead master agent for a long time from version 5.2.1. My system is configured as follows (embedded linux, kernel 2.4.19-rmk7-pxa1): - master agent (included modules: agentx/master; excluded: mibII ucd_snmp agent_mibs) - seven subagents (included modules: agentx/subagent mibII/vacm_vars and_some_proprietary_modules"; excluded: mibII agent_mibs) Subagents are dynamicaly executed-forked from external proces and killed with SIGTERM signal. When agent is booting up, registers corresponding row(s) in different tables and unregisters in case of shuting down. Registering is done by xxx_register_handler_xxx, xxx_register_instance_xxx, etc. In main thread trap handling routine is added via xxx_register_fd_xxx. So every subagent/master agent is running single threaded in separate process, only the master is the one that actually accepts snmp requests in send out traps. System was commenced on netsnmp 5.2.1.rc2. Everything works ok... except that after uncertain number of starting up and terminating subagents, responding to get/set requests, MASTER GOES DEAD. It is not blocked, it still forwards trap, if some is introduced to him via subagent, but DOES NOT RESPOND to any get/set request. UDP port 161 remains opened, memory footprint is the same, any subagent restarted, normaly registers its sessions, master still responds to signals (SIGHUP,...) I compiled the same configuration on i386 system, but hw configuration is not the same, so I could never reproduce the same dead situation. Also, everything lies on the embedded system, debuging is hard to achive... Few weeks ago I discovered, that all that seven subagent makes the persistent file, all with the same filename in the same directory. I changed that, that dynamicaly, at booting time, application name (previous snmpd) is rearanged in some unique name. And guess what: THERE IS ALMOST NO MORE DEAD master agent. Almost. If some request is pending to the master and consequently to the subagent, and in this moment the same agent receives SIGTERM signal, problem arise again. Master goes dead. Of course, if I restart master agent in this dark ages :), all agentx sessions are restored and system works as nothing happened at all. I found in mailing list some similar problem for 5.1.1 version and hoped, that my yesterday's upgrade to 5.2.2 will fade out feature of dead master agent. But it was not and I run out of ideas... Have somebody any idea, what is ...ing wrong? Happy christmans to everybody, Igor Igor Grašič, Ms. D. Development Engineer Iskra Transmission, d. d., Ljubljana Stegne 11, SI-1000 Ljubljana, Slovenia phone: +386 1 5003 000 fax: +386 1 5003 500 www.iskratr.si, www.iskratr.com ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! http://ads.osdn.com/?ad_idv37&alloc_id865&op=click _______________________________________________ Net-snmp-coders mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
