Synchrony Hazelcast Port Conflict Causing Collaborative Editing Failures

Platform Notice: Data Center Only - This article only applies to Atlassian apps on the Data Center platform.

Note that this KB was created for the Data Center version of the product. Data Center KBs for non-Data-Center-specific features may also work for Server versions of the product, however they have not been tested. Support for Server* products ended on February 15th 2024. If you are running a Server product, you can visit the Atlassian Server end of support announcement to review your migration options.

*Except Fisheye and Crucible

Summary

Users were unable to edit pages collaboratively and saw a loading delay message. Synchrony logs on the affected node showed that Hazelcast was not active and reported a bind error for a port conflict. A stale Synchrony process was still holding Hazelcast port 5701 during a rolling start, which prevented the new Synchrony instance from forming its Hazelcast cluster.

Environment

  • Confluence Data Center

Diagnosis

  • User popup on edit: “This page is taking longer to load than usual. Give it a few moments, then try refreshing. Still having issues? Contact your Confluence admin”

  • Collaborative editing unavailable

Evidence

Key log excerpts collected during the incident from atlassian-synchrony.log

2026-03-11 16:28:54,459 WARN [async-dispatch-100] [synchrony.sync.hubs] Hazelcast instance is not active! {...} :cause "Hazelcast instance is not active!" ..... 2026-03-11 16:37:02,826 ERROR [main] [synchrony.logging] uncaught exception {:throwable #error { :cause "Address already in use" :via [{:type clojure.lang.ExceptionInfo ...} {:type java.net.BindException :message "Address already in use" ...}] ...} 2026-03-11 16:48:33,174 INFO [TcpServerConnector] [...] Could not connect to: /xxx.xxx.xx.xx:5701. Reason: IOException[Connection refused ...] ......

Cause

During a rolling start, the existing Synchrony process on node didn’t shut down cleanly. Confluence started a new Synchrony process on the node while the stale process was still running. The stale process continued to hold Hazelcast port 5701, so the new Synchrony instance couldn’t bind to the port or initialize its Hazelcast cluster. This caused “Hazelcast instance is not active!” and “Address already in use” errors and disabled collaborative editing.

Solution

  • Stop Confluence on the affected node.

  • Identify and terminate the stale Synchrony/Confluence Java process that is still listening on Hazelcast port 5701.

  • Restart Confluence and confirmed that only a single Synchrony instance started.

  • Verify that “Hazelcast instance is not active!” messages no longer appeared in the Synchrony logs and that collaborative editing is working as expected.

Prevention and Best Practices

  • Before maintenance or rolling restarts, cleanly stop Confluence and confirm no Java/Synchrony processes remain.

  • Ensure required ports are free prior to startup: 5701 (Synchrony Hazelcast), 8091 (Synchrony HTTP).

  • Leverage commands or orchestration/monitoring to detect and clean up stale processes holding critical ports.

Verification Checklist

  • Single Synchrony process per node after restart

  • No bind errors for ports 5701/8091 in Synchrony logs

  • No new “Hazelcast instance is not active!” messages

Related Knowledge Base

Updated on March 20, 2026

Still need help?

The Atlassian Community is here for you.