Docker not starting after install OneAgent

[radek_jasinski]

Hi,

I have a big problem on one of my our environment.

After installing the oneagent on a host with OpenShift version 3.9 installed, we've a problem with Docker. All processes start correctly but only Docker has errors in logs and don't started.

Error starting daemon: Error initializing network controller: error obtaining controller instance: failed to create NAT chain: Could not create nat/DOCKER chain: ERROR: ld.so: object '/opt/dynatrace/oneagent/agent/processagent/lib64/liboneagentproc.so' from /etc/ld.so.preload cannot be preloaded: ignored.

Both files exist. Docker and oneagent are run on the root account.

/The installation is on the Centos Atomic./

Deployments:

● ostree://centos-atomic-host:centos-atomic-host/7/x86_64/standard

Version: 7.1811 (2018-12-06 22:08:05)

Commit: ee5a6f25f64d140a15a94278790840c3d58585b69dab9a1c33928d3cfe45cfd9

GPGSignature: Valid signature by 64E3E7558572B59A319452AAF17E745691BA8335

I have opened ticket for this in error, but maybe someone will help me in the meantime because the topic is urgent.

Regards,

Radek

[Julius_Loman]

• Selinux is enabled or not?
• What is your oneagent version?
• Is it a standard agent deployed on the host, or using image? (I assume it's locally deployed)
• What is the content of /etc/ld.so.preload?
• The path to library '/opt/dynatrace/oneagent/agent/processagent/lib64/liboneagentproc.so' is definitely non-standard. Does this file exist on your host? Default location should be /opt/dynatrace/oneagent/agent/lib64/liboneagentproc.so
  
  Although I have not experience with CentOS atomic, I don't see a reason why the location should have changed.

[radek_jasinski · Accepted solution]

Hi @Julius L.

I have information from support that this problem will be fixed in upcoming releases:

• The issue with Docker has been fixed with agent 1.159.
  
• Once Managed 1.160 is released(ETA end of January)

Radek

[Julius_Loman]

Do you know the cause of this issue? Is it somehow related to docker configuration?

[Tomasz_Gajger · Accepted solution]

Root cause was a change in default SELinux policy, most likely introduced in recent release of CentOS 7.6 Atomic.

SELinux module shipped with OneAgent was adjusted to account for these changes.