[IOTDB-1424] [Distributed] Clean inactive queries

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Core/Cluster
• Labels: keepalive query resource-management
• Assignee: Unassigned
• Reporter: Tian Jiang
• Created: 07/Jun/21 02:40
• Updated: 07/Jun/21 02:40

Description

In the standalone version, each query is bound to a sessionId, which is further bound to a connection. Since a connection has its own timeout and keepalive, there should not be any hanging sessionIds and thus no hanging queries.

But in the distributed version, queries may be issued from other nodes through general-purpose connections, and the requests to end such queries may be lost due to various reasons. As a result, internal queries (queries issued by other cluster nodes) may be hanging forever if the endQuery request is lost or even never sent because the query coordinator node crashes.

So for internal queries, a proper keepalive mechanism should be proposed to avoid resource leaking when the queries cannot be ended normally.