Handling RMI Stub Warnings in GemFire: Understanding and Gracefully Managing Locator Shutdowns
Article ID: 397717
Updated On:
Products
Issue/Introduction
Stopping a locator in GemFire—particularly one acting as the cluster coordinator—can lead to runtime warnings like below and cluster instability if not handle
This article describes the reason behind this exception, the role of the coordinator in GemFire clusters, and the best practices for shutting down cluster components gracefully to maintain stability and avoid communication failures.
Cause
This warning appears when an operation attempts to communicate with a locator whose RMI services are no longer available. Specifically:
-
The warning typically occurs when a locator acting as the coordinator is stopped.
-
Other members in the cluster, expecting the coordinator to still be available, try to communicate with it via RMI, resulting in a
NoSuchObjectException. -
If only one locator is present and it is shut down, the cluster loses coordination, leading to various exceptions and failures in cluster management.
By contrast, if the stopped locator is not the coordinator, the cluster can continue functioning using other active locators, and this warning does not appear.
Resolution
To prevent such issues and ensure cluster stability, it is essential to follow graceful shutdown procedures using the tools provided by GemFire, specifically gfsh (GemFire Shell).
- Use gfsh for shutdown the entire cluster stops all members in the cluster, ensuring that dependencies and state transitions are properly handled.
shutdown [--time-out=value] [--include-locators=value]
- Avoid stopping the only locator—especially if it is the coordinator—should be avoided unless absolutely necessary. Doing so causes:
-
-
Loss of RMI services
-
Inability to run
gfshcommands -
Risk of server disconnection or data unavailability
-
Feedback
