You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Chesnay Schepler (Jira)" <ji...@apache.org> on 2020/11/05 13:03:00 UTC
[jira] [Created] (FLINK-20003) Improve slot report logging messages
Chesnay Schepler created FLINK-20003:
----------------------------------------
Summary: Improve slot report logging messages
Key: FLINK-20003
URL: https://issues.apache.org/jira/browse/FLINK-20003
Project: Flink
Issue Type: Improvement
Components: Runtime / Coordination
Reporter: Chesnay Schepler
Assignee: Chesnay Schepler
Fix For: 1.12.0
Slot reports that are received by the ResourceManager are logged.
Currently, such a message looks like this:
{code}
16381 [flink-akka.actor.default-dispatcher-2] DEBUG org.apache.flink.runtime.resourcemanager.slotmanager.DeclarativeSlotManager [] - Received slot report from instance aabc4f58e13c5038349292df30a00d35: SlotReport{slotsStatus=[SlotStatus{slotID=d006c062-d452-4aaa-bdd3-951fba1e80e5_0, resourceProfile=ResourceProfile{managedMemory=20.000mb (20971520 bytes), networkMemory=16.000mb (16777216 bytes)}, allocationID=d8d7358d4ba604393dee5d495ee94288, jobID=ced59189ae30c41896bfd86a0217668d}, SlotStatus{slotID=d006c062-d452-4aaa-bdd3-951fba1e80e5_1, resourceProfile=ResourceProfile{managedMemory=20.000mb (20971520 bytes), networkMemory=16.000mb (16777216 bytes)}, allocationID=f4ea20ff0924653bd3e524d3a594b8ed, jobID=ced59189ae30c41896bfd86a0217668d}, SlotStatus{slotID=d006c062-d452-4aaa-bdd3-951fba1e80e5_2, resourceProfile=ResourceProfile{managedMemory=20.000mb (20971520 bytes), networkMemory=16.000mb (16777216 bytes)}, allocationID=3e191837d45879ab59b45eedf7685235, jobID=ced59189ae30c41896bfd86a0217668d}, SlotStatus{slotID=d006c062-d452-4aaa-bdd3-951fba1e80e5_3, resourceProfile=ResourceProfile{managedMemory=20.000mb (20971520 bytes), networkMemory=16.000mb (16777216 bytes)}, allocationID=null, jobID=null}]}.
{code}
I propose 2 changes:
1) Invert the order in which the slot status fields are printed. The job and allocation IDs are usually the interesting parts, with the resource profile commonly just being a bunch of noise.
2) introduce line-breaks (up to a limit), so that you can tell at a glance what is going on, particularly when debugging tests.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)