Hi Oak community,
I am running Oak 2.4.0 with a DocumentNodeStore backed by MongoDB and a
FileDataStore on a filesystem volume.
The deployment runs inside containers (Docker/Kubernetes) and only a single
application instance is expected to be active at any given time. There is no
intentional clustering and there is never more than one Oak instance accessing
the repository concurrently.
While investigating repository metadata, I noticed that every application
restart appears to create a new entry in the clusterNodes collection.
Using oak-run:
java -jar oak-run-2.4.0.jar clusternodes --verbose <mongo-uri>
I can see a long history of cluster node IDs. More surprisingly, all of them
are reported as ACTIVE even though only one application instance currently
exists.
A simplified example looks like:
53 ACTIVE ...
54 ACTIVE ...
55 ACTIVE ...
56 ACTIVE ...
The most recent node corresponds to the currently running instance, but older
nodes are still reported as ACTIVE as well.
While searching for related information I found OAK-7316 ("Greedy
ClusterNodeInfo"), which discusses clusterId allocation and reuse in
dynamic/containerized environments. This made me wonder whether what I am
observing is expected behavior in Kubernetes-based deployments.
ă1-6acb6băă2-445303ă
My questions are:
1. In a deployment where only a single Oak instance exists, should the same
clusterId normally be reused across restarts?
2. Is it expected and safe to accumulate many clusterNodes entries over time?
3. Is it expected that previous clusterNodes are still reported as ACTIVE after
the application has been restarted multiple times?
4. Is there a recommended configuration or operational practice to ensure that
a single-instance deployment keeps using a single cluster node identity?
5. If multiple ACTIVE clusterNodes remain in the clusterNodes collection, does
this have any impact on repository correctness, background operations,
revisions, or maintenance tasks?
I would appreciate any guidance on whether this is normal behavior for
containerized deployments or whether it suggests that something is not being
handled as expected during node shutdown and restart.
Thanks in advance for your help.
Best regards,
Raffaele Gambelli
Cordiali saluti / Best regards,
Raffaele Gambelli
Application Architect
E [email protected]<mailto:[email protected]>
M +39 3371641888
[https://images.eu.signature365.com/b2uknhefs98nwdub/img_USamABop1MkXu35s/v114.jpg]
[CEGEKA]
Via Ettore Cristoni, 84
IT-40033 Bologna (IT), Italy
T +39 02 2544271
WWW.CEGEKA.COM<https://www.cegeka.com>
