Description of problem:

When we try to start the foreman.service, It triggers the startup of foreman.socket at the backend. 

During this startup, we get to see "TCP_NODELAY failed: Operation not supported" printed in the system logs which is harmless in nature but unexpected to see.



Version-Release number of selected component (if applicable):

Red Hat Satellite 6.9
Red Hat Satellite 6.10
Red Hat Satellite 6.11


How reproducible:

Always

Steps to Reproduce:

1. Install a Satellite 6.9\6.10\6.11

2. Try; 

  # for i in stop start; do systemctl $i foreman.socket; sleep 5; done

  or simply # systemctl restart foreman.socket
  
  or
  
  # foreman-maintain service stop
  # foreman-maintain service start

3. Monitor /var/log/messages or check the status of foreman.socket using systemctl


Actual results:

Jun 21 16:17:37 satellite.example.com systemd[1]: TCP_NODELAY failed: Operation not supported
Jun 21 16:17:37 satellite.example.com systemd[1]: Starting Foreman HTTP Server Accept Sockets.
Jun 21 16:17:37 satellite.example.com systemd[1]: Listening on Foreman HTTP Server Accept Sockets.


Expected results:

No "TCP_NODELAY failed: Operation not supported" message


Additional info:

The foreman.socket is actually a UNIX socket but it includes a TCP option in it i.e. "NoDelay=true" which is a TCP only option

# grep NoDelay /usr/share/foreman/extras/systemd/foreman.socket /usr/lib/systemd/system/foreman.socket
/usr/share/foreman/extras/systemd/foreman.socket:NoDelay=true
/usr/lib/systemd/system/foreman.socket:NoDelay=true

# man 7 tcp | grep TCP_NODELAY -A4 | tail -6
--
       TCP_NODELAY
              If  set,  disable  the  Nagle  algorithm.  This means that segments are always sent as soon as possible, even if there is only a small amount of data.
              When not set, data is buffered until there is a sufficient amount to send out, thereby avoiding the frequent sending of small packets,  which  results
              in  poor  utilization of the network.  This option is overridden by TCP_CORK; however, setting this option forces an explicit flush of pending output,
              even if TCP_CORK is currently set.


But since we explicily mention "NoDelay" to "True" , When the unix socket starts, It tries to honor that option but fails because TCP_NODELAY is tcp only option. 


Workaround:

* Edit /etc/systemd/system/foreman.socket.d/installer.conf and add "NoDelay=false" at the end of the file

# cat /etc/systemd/system/foreman.socket.d/installer.conf
[Socket]
ListenStream=
ListenStream=/run/foreman.sock
SocketUser=apache
SocketMode=0600
NoDelay=false

* systemctl daemon-reload

* systemctl restart foreman.socket


Unless, my explanation is wrong or missing some important part and due to some reason, NoDelay should always be set to true in the case of foreman.socket,  I will suggest having this properly handled by satellite-installer and set "NoDelay=false" in /etc/systemd/system/foreman.socket.d/installer.conf .