IDM server upgrade

Latest response

Hi,

I had to upgrade IDM servers from RHEL 7.5 to RHEL 7.9 and I've encountered some issues which I want to share so others can avoid headaches.

So if you check the documentation it says that you just need to run

yum update ipa-*

and it will do the trick. Once it is done you can update the rest of the system packages.
This method works up to RHEL 7.7 and ipa-server-4.6.5-11.el7_7.4.x86_64

The first issue happens when you try to update the ipa-* packages from ipa-server-4.6.5-11.el7_7.4.x86_64 to ipa-server-4.6.6-11.el7.x86_64.
The actual yum command finishes successfully. Also, the ipa-server-upgrade command is successful which runs after the cleanups.
But once the command returns the replication is completely broken across all nodes. The reason is that the LDAP server is non-responsive.

Reason:
The upgrade requires the update of 389-ds-base and 389-ds-base-libs before running ipa-server-upgrade. Unfortunately, the ipa-server packages have no requirements for these 2 packages.

To fix this:
- ipactl stop & (it will hang at 'Stopping Directory server' just wait until it reaching that step)
- kill -4 $(ps -C ns-slapd -o pid=)
- wait until replication recovered between all active nodes
- ipactl start

To avoid:
I didn't test it on RHEL 7.7 but it was the fix when upgrading from RHEL 7.8 to RHEL 7.9 so I assume it helps here as well since the ipa-server-upgrade command restarts the directory service multiple times.
Simply run

yum upgrade

on the IDM server. It will update the 389-ds packages first then run the ipa-server-upgrade after yum's cleanup is done. After the replication is recovered on all nodes you can restart the server.

The second issue occurs when upgrading from RHEL 7.8 to RHEL 7.9.
The 'yum update ipa-*' will hang after yum's cleanup is done and the post actions are running. Replication is down across all nodes.

Reason:
Very similar. 389-ds-base and 389-ds-base-libs needs to be updated before ipa-server-upgrade run.
In the background you can find an 'ipactl restart' command hanging because the LDAP server is non-responsive.

To fix:
- kill -4 $(ps -C ns-slapd -o pid=)
- let yum finish the rest

To avoid:
Simply run

yum upgrade

This will finish without issues. After the replication is recovered on all nodes you can restart the server.

I hope it'll help others to have a smooth upgrade.

Responses