[HDFS-16535] SlotReleaser should reuse the domain socket based on socket paths

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.3.1, 3.4.0
• Fix Version/s: 3.4.0, 3.3.3
• Component/s: hdfs-client
• Labels: pull-request-available
• Assignee: Unassigned
• Reporter: Quanlong Huang
• Created: 11/Apr/22 03:35
• Updated: 01/Jun/23 08:35
• Resolved: 18/Apr/22 02:38

Description

HDFS-13639 improves the performance of short-circuit shm slot releasing by reusing the domain socket that the client previously used to send release request to the DataNode.

This is good when there are only one DataNode locates with the client (truth in most of the production environment). However, if we launch multiple DataNodes on a machine (usually for testing, e.g. Impala's end-to-end tests), the request could be sent to the wrong DataNode. See an example in IMPALA-11234.

Issue Links

• is caused by: HDFS-13639 SlotReleaser is not fast enough
• is required by: IMPALA-11234 impalad keeps reporting ShortCircuitCache slot release failures in heavy workload
• links to: GitHub Pull Request #4158