Crashes due to invalid URL when Nodeport is pointing at pod

[capnajax]

This is on Kubernetes

1. Create a pod (I used a deployment in my test) that runs this container and expose port 389. Pod runs ok.
2. Create a NodePort service to access this container and expose 389 as 30389. Pod still running ok
3. Restart the pod. It crashes shortly after startup

*** /container/run/process/slapd/run started as PID 520
5f24e641 @(#) $OpenLDAP: slapd 2.4.50+dfsg-1~bpo10+1 (May  4 2020 05:25:06) $
	Debian OpenLDAP Maintainers <[email protected]>
5f24e641 daemon: listen URL "ldap://ldap-565f78fdb8-9x8nn:tcp://10.108.253.216:389" parse error=5
5f24e641 slapd stopped.
5f24e641 connections_destroy: nothing to destroy.
*** /container/run/process/slapd/run exited with status 1
*** Run commands before finish...
*** Killing all processes...

Note the URL is ldap://<pod name>:tcp:<container IP>:389 -- looks like there should have been a space in there.

[zerkms]

It happens because you named your service LDAP and kubernetes automatically creates LDAP_PORT and number of other environment variables.

And LDAP_PORT collides with this docker image conventions.

[SerialVelocity]

Ugh, this is because of #403. This will probably affect quite a lot of people when upgrading.

For Kubernetes you can set enableServiceLinks: false in the pod spec to get around this. Much better than trying to rename the service and updating all applications using ldap.