Description

This KB describes a scenario when number of stale sockets for eventd process on the master RE increases after each crash of the standby RE.

Symptoms

If the issue is hit, number of sockets opened by eventd process on the master RE will increase every time the standby RE crashes. These sockets can be checked using the following shell commands on the master RE:

ss -tonp src 128.0.0.4:7209 dst 128.0.0.5

For further details, please refer to https://prsearch.juniper.net/PR1783320

Solution

Fix for https://prsearch.juniper.net/PR1783320 added keepalives to TCP sockets used by eventd process to allow clean-up of stale sockets, please refer to this PR for the list of releases with the fix.

Sample output from a fixed release: 
[vrf:none] root@device:~# ss -tonp src 128.0.0.4:7209
State      Recv-Q       Send-Q             Local Address:Port             Peer Address:Port       Process
ESTAB      0            0                  128.0.0.4%iri:7209               128.0.0.16:54356       users:(("eventd",pid=15147,fd=41)) timer:(keepalive,1min28sec,0)
ESTAB      0            0                  128.0.0.4%iri:7209                128.0.0.5:45888       users:(("eventd",pid=15147,fd=38)) timer:(keepalive,1min12sec,0)
ESTAB      0            0                  128.0.0.4%iri:7209                128.0.0.4:39302       users:(("eventd",pid=15147,fd=35)) timer:(keepalive,55sec,0)

Modification History

2024-04-24 : Article Created