Jordan Hatcher created FLINK-15043:
--------------------------------------
Summary: Possible thread leak in task manager
Key: FLINK-15043
URL:
https://issues.apache.org/jira/browse/FLINK-15043 Project: Flink
Issue Type: Bug
Components: Runtime / Task
Affects Versions: 1.8.2
Reporter: Jordan Hatcher
We have a few flink clusters running version 1.8.2, with a mongo 4.2 database. We initially noticed a large number (16000-19000) of connections to the database, which we traced back to the flink task manager. Restarting the task managers dropped this number down to a few hundred connections. The number of database connections roughly corresponds to the number of task manager threads as reported by the metrics API (/taskmanagers/metrics?get=Status.JVM.Threads.Count) across a few different clusters, so we suspect there is some kind of thread leak occurring in the task manager. Running a few jobs through our flink workflow shows that the number of database connections is increasing by about 50 for each job we run.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)