[IMPALA-6196] Track per-remote-host RPC metrics

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Component/s: Backend, Distributed Exec
• Labels: observability
• Epic Link: Improve Hang Diagnostics
• Epic Color: ghx-label-2
• Assignee: Unassigned
• Reporter: Lars Volker
• Created: 16/Nov/17 00:19
• Updated: 23/Dec/20 18:18
• Resolved: 23/Dec/20 18:18

Description

We should track RPC statistics (latency, throughput) per remote host on each backend of a cluster. That can help identify network contention and disruptions. It could also be used during planning / fragment instance scheduling.

Issue Links

• duplicates: IMPALA-9235 Backport Kudu socket stats for /rpcz
• relates to: IMPALA-6088 Rack aware broadcast operator