Bug 265798 – net-analyzer/net-snmp - snmpd tries to monitor disconnected ppp interfaces

• Product: Gentoo Linux
• Status: RESOLVED UPSTREAM

[Description NiTr0 2009-04-11 18:49:52 UTC]

I have an PPTP server that assigns for server-side tunnel IP one address for all clients (for ex., 10.0.0.1). But after some time SNMP grows, takes a lot of memory, and a lot of CPU time when receive query, also - it returns on query a lot of duplicated interfaces:

# snmpwalk -v 3 -l AuthNoPriv -u User -A Pass 192.168.0.1 .1.3.6.1.2.1.2.2.1.2 -a SHA
....................
IF-MIB::ifDescr.1739 = STRING: ppp87
IF-MIB::ifDescr.1740 = STRING: ppp51
IF-MIB::ifDescr.1742 = STRING: ppp51
IF-MIB::ifDescr.1750 = STRING: ppp51
IF-MIB::ifDescr.1751 = STRING: ppp152
....................

It is actual for many (possible all?) versions of SNMP (before I saw this trouble on CentOS 5 & Fedora 8 -  5.3.1 and 5.4.1).

[Comment 1 Jeroen Roovers (RETIRED) 2009-04-14 15:48:07 UTC]

Please post your `emerge --info' too.

[Comment 2 Jeroen Roovers (RETIRED) 2009-04-14 15:48:23 UTC]

Oh right, all versions...

[Comment 3 NiTr0 2009-06-12 22:06:57 UTC]

Created attachment 194458 [details, diff]
Adapted patch from http://sourceforge.net/tracker/index.php?func=detail&aid=1513191&group_id=12694&atid=312694

There is a patch on sourceforge net-snmp bugtracker that removes dropped interfaces from statistics. As I understand, interfaces are removed after 5 minute absent.

I made "port" of it to current net-snmp-5.4.2.1 (it fails in 2nd hunk - in current SVN version there is an int, not an u_long variable 'lastchange'), and now it's in testing on server - at 1st look, it works correctly.

[Comment 4 NiTr0 2009-08-31 12:12:42 UTC]

Patched snmpd looks stable. IMHO it'll be good to include it into portage tree.

[Comment 5 Diego Elio Pettenò (RETIRED) 2012-08-19 02:35:26 UTC]

Please report this upstream.