The MMP Uses SO_LINGER To Cause TCP Rst, Runs Afoul Of Solaris Rst Rate Limit (Doc ID 2373751.1)

Last updated on MARCH 31, 2022

Applies to:

Symptoms

Using:  Oracle Communications Messaging Server 8.0.2.2.20180130

Running into the default for these Solaris ndd parameters:

tcp_rst_sent_rate_enabled = 1
tcp_rst_sent_rate = 40

You can see how many TCP Rst are NOT sent because of that rate limit by doing:

kstat -p tcp:0:tcpstat:tcp_rst_unsent

Looking at snoop, the Rst packets seem to be going to port 993 on the backends.

If we have a high enough rate of connections between the MMP and the backends, and if the MMP normally sets SO_LINGER such that connections are closed with Rst, it could be a normal connection rate is running afoul of the default for that ndd limit.

Changes

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.