Jira Data Center REST API thread persistence after abrupt client disconnect

Answer to the question

Jira Data Center does not automatically terminate the background worker thread that processes a long‑running REST operation when the client connection is lost. The thread continues to run until the operation completes or the database transaction times out.

Confirmed facts (based on Atlassian documentation and community evidence)

• The jira.rest.http.connectionTimeout setting controls only the HTTP client side; it does not influence the lifecycle of server‑side worker threads.
• Long‑running operations such as bulk changes, large exports, or custom background jobs are executed in dedicated worker threads that persist independently of the HTTP request.
• There is no built‑in configuration property that maps an HTTP disconnect or servlet container timeout to the cancellation of the underlying background task.
• The only automatic cleanup is the database transaction timeout, which may release locks but does not stop the Java thread.
• Thread dumps and JMX metrics confirm that the worker thread remains alive after a client disconnect.
• Manual cancellation can be performed through the Jira UI (e.g., “Cancel bulk change”) or by implementing custom plugin logic that monitors and aborts the task.

Why this matters

Orphaned worker threads hold database locks and consume JVM resources, potentially leading to thread exhaustion under high load. Understanding that Jira does not cancel them automatically helps you design mitigation strategies.

How to verify the behavior in your environment

1. Start a known long‑running REST operation (e.g., bulk change via the REST API).
2. Abruptly terminate the client (kill the process or drop the connection).
3. Within a few seconds, capture a thread dump (e.g., jstack -l > dump.txt) and look for the thread that started the bulk change. It should still be in a RUNNABLE state.
4. Check Jira logs for entries such as Connection reset or Client disconnected and verify that the bulk change continues in the logs after the disconnect.
5. Optionally, use JMX or the Jira admin console to monitor the number of active worker threads before and after the disconnect.

Mitigation options

• Manual cancellation: If the operation is still visible in the Jira UI, use the “Cancel” button. For REST‑initiated tasks, the UI may expose a cancellation endpoint in newer releases.
• Custom plugin: Write a plugin that registers a Listener on the relevant event (e.g., BulkChangeEvent) and calls TaskManager.cancelTask(taskId) when a client disconnect is detected.
• Database timeout tuning: Increase the database transaction timeout to a reasonable value that balances performance and resource cleanup.
• Resource monitoring: Use JMX or a monitoring tool to alert when the number of active worker threads exceeds a threshold.

What you might need to know next

If you are running a very recent Jira Data Center version (e.g., 8.19+), check the release notes for any new task cancellation APIs. If you discover a property such as jira.rest.abortOnClientDisconnect in your configuration, it would change this recommendation, but such a property does not exist in the current documentation.

Next steps for you

Run the verification steps above in your environment. If you observe that worker threads are not being cleaned up after a client disconnect, consider implementing one of the mitigation options. If you find a configuration that changes this behavior, let us know so we can update the guidance.