bsnmpd resource exhaustion issue?

[pboehmer]

Before I file a PR, has anyone else noticed issues with bsnmpd exhausting system resources to the point of lock up?

I noticed a number of recently updated (~past 3 weeks) 10.3-STABLE boxes started showing bsnmpd high memory/cpu usage in top when the /usr/lib/snmp_hostres.so is enabled.

Initially, I thought the daemon was being "over polled" remotely, but I have blocked all access to the SNMP ports via /sbin/ipfw (including localhost) to isolate the issue with no change in outcome. I am using the default config with changes only to the READ community string, location, and contact options in addition to enabling the /usr/lib/snmp_hostres.so option.

 

[xaa]

Have you tried looking with ktrace / kdump what it was doing?

I had a similar issue today after upgrading yesterday. With ktrace I found it was looking at the /dev/cd0 with a message "no medium" over and over again. So I inserted a CD in the drive and snmpd went back to normal.

Looks like something is overly edgy on checking the error messages there?