SNMPD Consuming Swap Space

[Lowell Gilbert]

"Davenport, Steve M" <SDavenpo at mc.utmck.edu> writes:

> Hello,
>  
> I am running snmpd 5.4.1.2 built from a port on 7-Release, hardware is
> Sun V100, 512Mb total memory, 381Mb free. The daemon starts fine. When a
> snmpwalk is done from another system all is well until the interface
> table and then I see:
>  
> interfaces.ifTable.ifEntry.ifSpeed.1 : Gauge32: 10000000
> interfaces.ifTable.ifEntry.ifSpeed.2 : Gauge32: 0
> interfaces.ifTable.ifEntry.ifSpeed.3 : Gauge32: 0
> snmpwalk: No response arrived before timeout.
>
> After the timeout happens, looking at "swapinfo -k" shows that swap
> space is continually consumed until empty at which point the snmpd
> daemon is stopped:
>  
> fbsdh# swap_pager: out of swap space
> swap_pager_getswapspace(16): failed
> Nov 25 11:44:08 fbsdh kernel: pid 84674 (snmpd), uid 0, was killed: out
> of swap space
>  
> Now the swap space is freeded:
> fbsdh# swapinfo -k
> Device          1K-blocks     Used    Avail Capacity
> /dev/ad0b         2099160    26456  2072704     1%
>  
> Has anyone seen this issue or know of a solution. 

Since you're running the SNMP daemon from a port, not the one from the
base system, you may need developers' assistance for this sort of bug.
Problems like this are often caused by wraparound bugs or infinite loops
in my experience, although there are lots of other possibilities.  If
you can get a stack traceback when it crashes, that will is quite likely
to give a solid clue.

-- 
Lowell Gilbert, embedded/networking software engineer, Boston area
		http://be-well.ilk.org/~lowell/