Upgrade from 2.11.0 to 3.3.1 failing agent/opensearch-agent.jar

Versions (relevant - OpenSearch/Dashboard/Server OS/Browser):3.3.1

Describe the issue:Recently, we upgraded our OpenSearch version to 3.3.1.
During the upgrade from 2.11.0 to 3.3.1, the upgrade process completes successfully, but we are observing pod restarts caused by the following error:

Error opening zip file or JAR manifest missing: agent/opensearch-agent.jar

  • The existing ingest pod (version 2.11.1) is already running in a 3/3 Ready state.

  • When we trigger the upgrade, a new ingest pod (version 3.1.1) is created.

  • It usually takes around 3–4 minutes for the new pod to reach the 3/3 Ready state.

  • During this time, the old 2.11.1 ingest pod continues running in 3/3 Ready state, so there is no service impact.

  • Once the new 3.1.1 pod becomes healthy, the old 2.11.1 pod is terminated, traffic switches seamlessly to the new pod, and the upgrade completes normally.

However, in the problematic case:

  • While the new ingest 3.1.1 pod is still initializing (still in its 3–4 mins to start),

  • The old ingest 2.11.1 pod unexpectedly restarts.

  • This results in no active ingest pod during that period(Zero downtime exists).

  • Because of this, there is traffic loss, and we observed the mentioned error in the logs of the restarted pod.

My question is: why are the older version pods restarting in the first place?

Configuration: provided 3 ingest 3 master and 3 data pods

Relevant Logs or Screenshots:

@Badamkoushil How did you deploy/upgraded the cluster? Did you use helm charts?
If so, could you share your values.yml file?

Hi @pablo , Actually, regarding the same query — the failing agent JAR is only a suspected cause. The main point is that the OpenSearch upgrade from 2.11.0 to 3.1.1 is not a supported upgrade path. Because of this, during the upgrade the cluster tries to resync, which leads to downtime.

To avoid this downtime, do we have any possible workaround during the upgrade process?

TIA/Ashok

Hi @pablo , as @AshokPonna mentioned earlier, this is only a suspicion — the main focus is the OpenSearch upgrade from 2.11.0 to 3.3.1.

To answer your question: we extract the OpenSearch tar file and run it as a binary inside our Kubernetes pods. During the upgrade, we replace the existing pod with a new Kubernetes pod that contains the updated OpenSearch folder (version 3.3.1).

Regards,

Koushil

@Badamkoushil @AshokPonna What do you mean that 2.11.0 to 3.3.1 is not supported?
According to OpenSearch documentation, upgrade between major versions i.e. 2.x and 3.x or 1.x and 2.x is supported.

Regarding the upgrade process. @Badamkoushil This looks very manual and I understand that you don’t use official OpenSearch Helm Charts or OpenSearch Operator.

Could you describe your deployment in more details? How many nodes do you have?
Could you share output of this command?

GET _cat/nodes 

OpenSearch pods only

kubectl get pods 

Did you use deployment or stateful sets to create OpenSearch cluster?

Hi @pablo , Thanks for your response, but when we are trying to upgrade from 2.11.1 to 3.3.1, upgrade is working but with some downtime, between this time we could see error logs as below.
{“extra_data”:{“class”:“c.e.b.a.s.o.p.s.t.TlsTransport”},“message”:“Peer not connected [channel: Netty4TcpChannel{localAddress=/192.168.161.67:9300, remoteAddress=/192.168.52.103:50358}] Cause: Received message from unsupported version: [2.11.1] minimal compatible version is: [2.19.0]”,“metadata”:{“container_name”:“master”,“namespace”:“ashok”,“pod_name”:“opensearch-master-2”},“service_id”:“opensearch”,“severity”:“warning”,“timestamp”:“2025-11-20T12:15:45.180+00:00”,“version”:“1.2.0”}
{“extra_data”:{“class”:“c.e.b.a.s.o.p.s.t.TlsTransport”},“message”:“Exception caught onException TlsTransport cause : Peer not connected [channel: Netty4TcpChannel{localAddress=/192.168.161.67:9300, remoteAddress=/192.168.52.103:50358}] Cause: Received message from unsupported version: [2.11.1] minimal compatible version is: [2.19.0]”,“metadata”:{“container_name”:“master”,“namespace”:“ashok”,“pod_name”:“opensearch-master-2”},“service_id”:“opensearch”,“severity”:“warning”,“timestamp”:“2025-11-20T12:15:45.180+00:00”,“version”:“1.2.0”}
{“extra_data”:{“class”:“c.e.b.a.s.o.p.s.t.TlsTransport”},“message”:“Peer not connected [channel: Netty4TcpChannel{localAddress=/[fc00:1000:0:0:0:0:0:a144]:9300, remoteAddress=/[fc00:1000:0:0:0:0:0:3444]:48842}] Cause: Received message from unsupported version: [2.11.1] minimal compatible version is: [2.19.0]”,“metadata”:{“container_name”:“master”,“namespace”:“ashok”,“pod_name”:“opensearch-master-2”},“service_id”:“opensearch”,“severity”:“warning”,“timestamp”:“2025-11-20T12:15:45.181+00:00”,“version”:“1.2.0”}
{“extra_data”:{“class”:“c.e.b.a.s.o.p.s.t.TlsTransport”},“message”:“Exception caught onException TlsTransport cause : Peer not connected [channel: Netty4TcpChannel{localAddress=/[fc00:1000:0:0:0:0:0:a144]:9300, remoteAddress=/[fc00:1000:0:0:0:0:0:3444]:48842}] Cause: Received message from unsupported version: [2.11.1] minimal compatible version is: [2.19.0]”,“metadata”:{“container_name”:“master”,“namespace”:“ashok”,“pod_name”:“opensearch-master-2”},“service_id”:“opensearch”,“severity”:“warning”,“timestamp”:“2025-11-20T12:15:45.182+00:00”,“version”:“1.2.0”}

BR/Ashok

@AshokPonna This is a statement from the AWS OpenSearch managed service

If you’re running OpenSearch 1.3 or 2.x, you must first upgrade to OpenSearch 2.19 before upgrading to OpenSearch 3.1.

This confirms your finding.

I’ve raised a GitHub issue to update the documentation.