This is a RHEL + CentOS issue introduced by how the connector to systemd works.

The effect is well described here:

I can't make anything of the comment 

The contents of /var/lib/rsyslog/imjournal.state are only manipulated through systemd API.

I mean, we care if our syslog is working, or not. Not whose manipulating the file to a way that it'll later be unable to read it. Besides, reading of the PR shows any event that sets this file to 0 bytes will fry it.

The other case is if it has a wrong time pointer after a time correction: it will stop working again.

You'll find a corresponing message in /var/log/messages:

Sep 19 19:31:38 dhcp204 rsyslogd: [origin software="rsyslogd" swVersion="7.4.7" x-pid="32097" x-info=""] start
Sep 19 19:31:38 dhcp204 rsyslogd-2307: warning: ~ action is deprecated, consider using the 'stop' statement instead [try ]
Sep 19 19:31:38 dhcp204 rsyslogd-2027: imjournal: fscanf on state file `/var/lib/rsyslog/imjournal.state' failed
 [try ]
Sep 19 19:31:58 dhcp204 rsyslogd: [origin software="rsyslogd" swVersion="7.4.7" x-pid="32097" x-info=""] exiting on signal 15.

And basically that'll be all it's logging!

The 2027 error is a generic file IO alert:

Makes sense when it only gets to read 0 bytes of that file, right?

The more detailed story is in this bug report:


To verify

Check this:

  1. You have an existing /var/lib/rsyslog/imjournal.state and it's possibly 0 bytes
  2. You see your remote log only being used when you restart rsyslog, and all you see is the error messages. No other log traffic.

tcpdump example

tcpdump -nn -vv -i ens3 port 514



To fix

I've used the following script:

systemctl restart rsyslog 
if journalctl --since="-1m" | grep -q ; then
  systemctl stop rsyslog 
  rm /var/lib/rsyslog/imjournal.state
  systemctl start rsyslog 



To monitor

I'm using a manual (not inventory based) check to alert if that file is ever created.

First, on the remote node in filewatch.cfg i'm setting up tracking of this file.

# cat /etc/check_mk/fileinfo.cfg 

This is accompanied by a manual check, meaning it's enforced no matter if inventory sees that file.

I've not yet added the condition to the file size.

It seems this file can be around and larger than 0 bytes and things might still work.


Finally I also added an error condition on the rsyslog error to logwatch.cfg:

# grep -e messages -e rsyslog /etc/check_mk/logwatch.cfg 
# C: Critical messages
# W: Warning messages
 C rsyslogd-2027: imjournal: fscanf:



more things...

I think this file should be in /var/spool/rsyslog, not in /var/lib/rsyslog

The whole design is a clusterfuck. If you have some RedHat support, please put pressure on this bug.

The rsyslog docs actually tell you to NOT USE the imstate module.

I've later run into a rate limiting issue

See for more info about this

And, for your amusement, this is what a fix looks like if you're using Rudder:

Which is no more than 22 lines of policy, including comments, and fixes even this issue.