HDDS-7739. EC: Increase the information in the RM sending command log message

[sodonnel]

What changes were proposed in this pull request?

Replication Manager writes an info log each time it sends a command to a datanode, but aside from the command name, it doesn't contain any further information about the command. This change is to log more details about the command being sent which may aid debugging issues in the future.

Example new messages after this change:

2023-01-06 17:23:06,080 [main] INFO  replication.ReplicationManager (ReplicationManager.java:sendDatanodeCommand(426)) - Sending command [replicateContainerCommand: containerId: 1, replicaIndex: 1, sourceNodes: [localhost-143.110.182.62/143.110.182.62, localhost-243.150.55.19/243.150.55.19]] for container ContainerInfo{id=#1, state=CLOSED, pipelineID=PipelineID=bdd96da4-f447-45d4-9331-62480e7bf224, stateEnterTime=1970-01-01T00:00:00Z, owner=Ozone} to f8bca9d9-cf71-4e8a-aaed-f48ddc148a73{ip: 37.218.105.163, host: localhost-37.218.105.163, ports: [STANDALONE=0, RATIS=0, REST=0, REPLICATION=0, RATIS_ADMIN=0, RATIS_SERVER=0, RATIS_DATASTREAM=0], networkLocation: /default-rack, certSerialId: null, persistedOpState: IN_SERVICE, persistedOpStateExpiryEpochSec: 0}

2023-01-06 17:23:05,893 [main] INFO  replication.ReplicationManager (ReplicationManager.java:sendDatanodeCommand(426)) - Sending command [closeContainerCommand: containerID: 1, pipelineID: PipelineID=6dad5a63-dfa1-4621-b8c4-7de81e024cdc, force: true] for container ContainerInfo{id=#1, state=CLOSED, pipelineID=PipelineID=6dad5a63-dfa1-4621-b8c4-7de81e024cdc, stateEnterTime=1970-01-01T00:00:00Z, owner=Ozone} to bb47be9c-0886-4852-8ca7-46799ac36883{ip: 140.122.156.42, host: localhost-140.122.156.42, ports: [STANDALONE=0, RATIS=0, REST=0, REPLICATION=0, RATIS_ADMIN=0, RATIS_SERVER=0, RATIS_DATASTREAM=0], networkLocation: /default-rack, certSerialId: null, persistedOpState: IN_SERVICE, persistedOpStateExpiryEpochSec: 0}


2023-01-06 17:23:05,938 [main] INFO  replication.ReplicationManager (ReplicationManager.java:sendDatanodeCommand(426)) - Sending command [reconstructECContainersCommand: containerID: 1, replicationConfig: EC/ECReplicationConfig{data=3, parity=2, ecChunkSize=1048576, codec=rs}, sources: [localhost-208.151.31.91/208.151.31.91 replicaIndex: 1, localhost-27.100.52.223/27.100.52.223 replicaIndex: 2, localhost-247.11.160.69/247.11.160.69 replicaIndex: 3], targets: [localhost-220.147.5.156/220.147.5.156, localhost-193.210.182.11/193.210.182.11], missingIndexes: [4, 5]] for container ContainerInfo{id=#1, state=CLOSED, pipelineID=PipelineID=874dd4ae-c222-4f91-b587-81551ef78a20, stateEnterTime=1970-01-01T00:00:00Z, owner=Ozone} to ecd0da3f-68b0-4bf8-85da-b7d3a41100a2{ip: 220.147.5.156, host: localhost-220.147.5.156, ports: [STANDALONE=0, RATIS=0, REST=0, REPLICATION=0, RATIS_ADMIN=0, RATIS_SERVER=0, RATIS_DATASTREAM=0], networkLocation: /default-rack, certSerialId: null, persistedOpState: IN_SERVICE, persistedOpStateExpiryEpochSec: 0}

2023-01-06 17:23:05,987 [main] INFO  replication.ReplicationManager (ReplicationManager.java:sendDatanodeCommand(426)) - Sending command [deleteContainerCommand: containerID: 1, replicaIndex: 1, force: false] for container ContainerInfo{id=#1, state=CLOSED, pipelineID=PipelineID=4c4b1a01-ebbf-466a-a1f3-45b306533709, stateEnterTime=1970-01-01T00:00:00Z, owner=Ozone} to 7c838ac2-d7bd-4719-ac38-f1ff5adb7869{ip: 122.2.174.81, host: localhost-122.2.174.81, ports: [STANDALONE=0, RATIS=0, REST=0, REPLICATION=0, RATIS_ADMIN=0, RATIS_SERVER=0, RATIS_DATASTREAM=0], networkLocation: /default-rack, certSerialId: null, persistedOpState: IN_SERVICE, persistedOpStateExpiryEpochSec: 0}

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-7739

How was this patch tested?

Logging only change. I ran some tests which caused the logs to be emitted and checked their format (posted above).

[adoroszlai]

I would love to make this (or UUID + this) the default toString, and renaming the existing more verbose toString to something like toDebugString. This way we could avoid having to convert manually in all places, and use the verbose one explicitly if needed.

[sodonnel]

I'm fine with adding UUID into the string too. It would solve the problem with mini-cluster where all the nodes are on the same host / IP. I wonder if anything would break if we changed the default datanode details to string? Almost everywhere it is logged, we don't call about all the verbose info it currently includes. We really just want to the host / IP / ID.