Description
Elasticsearch Version
7.14.2
Installed Plugins
No response
Java Version
openjdk version "1.8.0_232"
OS Version
CentOS Linux release 7.9 (Final)
Problem Description
When a query containing one hundred terms causes a memory meltdown in node a, which then cancels the tasks of other nodes, and causes Circuit breaker exception in two other nodes (b, c), running log like this:
[2025-01-22T11:38:51,422][WARN ][o.e.t.InboundHandler ] [xxx] Circuit breaker exception[transport_response], bytes wanted[30950235606], bytes limited[30893565542], status[TOO_MANY_REQUESTS]
The GC situation of node a is as follows：
many errors occurs at node b,c :
[2025-01-22T11:38:48,950][WARN ][o.e.t.TaskCancellationService] [xxx] failed to remove ban for tasks with the parent [xxx:4469158904] on connection [NodeChannels[{xxx}{jBzstAr8T_OdeyQa7K8lHw}{xl6G7wcfQCez-1kCF37Q6Q}{xxx}{xxx:9300}{hilmrst}]]: [xxx][xxx:9300][internal:admin/tasks/ban]
The GC situation of node b,c is as follows：
Steps to Reproduce
Why does a large number of cancel tasks cause the memory to rise so quickly, until leading to a memory Circuit breaker exception
Logs (if relevant)
No response