LLAWP process CPU usage goes to 100%+

Products

• CA Single Sign On Secure Proxy Server (SiteMinder)
• CA Single Sign On SOA Security Manager (SiteMinder)
• CA Single Sign-On

Issue/Introduction

- We have a new agent installation that is making the CPU usage climb to 100% + after some time. this is a production system and needs to be resolved right away. I tried to enable the trace logs from the LocalConfig.conf, but the web agent logs is not even showing any update happening on it, so it has become unresponsive.

 

Environment

- SPS 12.52 SP1 CR04

- RHEL 7.2/CentOS 7.2

Cause

- The issue was discovered to be a "feature" integrated into RHEL 7, where the semaphores are getting removed by OS after user logout from session.

 

Resolution

- With RHEL/CentOS 7.2, the system logind will now remove all IPC's (semaphores) for a user when logging out. For some applications this will cause issues. On CA Single Sign-On, this causes the CPU to spike to 100% while it tries to manage the missing IPCs (semaphores). 

- As a workaround, all RHEL/CentOS 7.2 or better servers that require CA Single Sign-On will make the following changes:

1. Create directory /usr/lib/systemd/logind.conf.d 

2. Create file /usr/lib/systemd/logind.conf.d/siteminder.conf 

3. Edit file /usr/lib/systemd/logind.conf.d/siteminder.conf to contain the following: 

4. [Login]

5. RemoveIPC=no

6. Restart the logind service - systemctl restart systemd-logind