#215 at Controller restart, behaviour has changed since 2.5.3

[Bruno]

Dear all,
My JS7 application is hosted on a kubernes cluster, since the new version 2.5.3, when i restart the Controller Pod, i have the following message (see picture), agent is no more in ok Status, and my workflow are nomore marked as synchronized/deployed.

Do i have to persist the Controller hostname ? (statefullset) and what about the Agent Host name ? it was not necessary for 2.5.1.

[Andreas]

Hi Bruno,

hostnames have to be persisted. You configure the URL of the Controller and Agents in JOC Cockpit. How should JOC Cockpit connect to the Controller and the Controller to Agents if hostnames (or IP addresses) are no longer valid?

This is not new for release 2.5.3.

The immediate problem is related but different. I'm not certain how comes that the same hostname is inserted twice into the inv_operating_systems table. If you can forward JOC Cockpit's jetty.log and joc.log files then this will shed more light. As a workaround you can remove the offending entry from the table.

Did you by chance perform the recent JS7 upgrade using a different security level when installing JOC Cockpit? This is visible in the JOC Cockpit dashboard from the JOC Cockpit icon and from the column security_level in table joc_instances.

Bests regards
Andreas

[Bruno]

Dear Andreas,
Thank you for your quick response .Url points to kubernetes services (see picture for agent url coupling failed)- they do not change . But if i delete the pod hosting the service, when the pod is created with a new name, the problem appears (and the message includes the previous pod name, not the service url). it seems to me that the behavior has changed. I cannot find any information about security level on the dashboard nor on the icon..

[Andreas]

Hi Bruno,
you should identify the security level from the dashboard, see screenshot.

By default the pod's name will be used as its hostname. And if you do not persist the hostname then you will have a new hostname each time that the prod is re-created.

This cannot work as the network connections between JOC Cockpit, Controller and Agents are based on hostnames. There is no change in this mechanism since the very first release of JS7.

Please, persist the hostnames of services in your pods.

Best regards
Andreas

[Bruno]

Dear Andreas, i've persisted hostname without change on the behaviour, BUT i've found what is the problem and solve it ! the path to persist state and logs has changed in this news version :
for controller previously it was /sos-berlin.com/js7/controller/var/state and now /sos-berlin.com/js7/controller/state
for agent previously it was /sos-berlin.com/js7/agent/var_4445/state and now /sos-berlin.com/js7/agent/state