Nikita Amelchev created IGNITE-28927:
----------------------------------------
Summary: Task name hash resolution fails after cluster
reactivation for long-running tasks
Key: IGNITE-28927
URL: https://issues.apache.org/jira/browse/IGNITE-28927
Project: Ignite
Issue Type: Bug
Reporter: Nikita Amelchev
{{GridTaskProcessor.resolveTaskName}} in Apache Ignite 2 resolves a task name
from its hash using the system cache.
A bug occurs in the following scenario:
# A long-running task is started. (Such tasks can continue running while the
cluster is inactive.)
# The cluster is deactivated, which clears the system cache.
# The cluster is activated again.
# The running task attempts to write to a cache.
At this point, task name resolution fails because the task name hash can no
longer be resolved—the system cache has been cleared during deactivation.
As a result, operations relying on {{GridTaskProcessor.resolveTaskName}} fail
for tasks that survive a cluster deactivation/activation cycle.
*Possible solution*
One possible approach is to store the task session ID instead of the task name
hash. This would eliminate the dependency on the system cache for task name
resolution. However, this shifts the responsibility of mapping a session ID
back to the corresponding task to the user, making diagnostics less convenient.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
