[HDDS-4538] SCM HA needs handle the generation of clusterID and scmUuid in a decent way.

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.1.0
• Component/s: SCM HA
• Labels: pull-request-available
• Assignee: Unassigned
• Reporter: Glen Geng
• Created: 02/Dec/20 07:42
• Updated: 19/Nov/24 18:59
• Resolved: 19/Nov/24 18:59

Description

During the initialization of SCM, it will generate a random clusterID and scmUuid, which will be sent to Datanode during their Version request.

 

On Datanode, The disk layout per volume is as follows:

../hdds/VERSION
../hdds/<<scmUuid>>/current/<<containerDir>>/<<containerID>>/metadata
../hdds/<<scmUuid>>/current/<<containerDir>>/<<containerID>>/<<dataDir>>

The metadata file for a typical container is as follows:

[ozoneadmin@tdw-9-179-144-104 /data6/hdds/hdds/326a5fe1-e63c-44b6-a57e-2f858fe4eaa7/current/containerDir0/262574/metadata]$ cat 262574.container 
!<KeyValueContainerData>
checksum: f1473bc5a9f9fa307edf2040edf0fb5ac40912a5bd8610e2d42fa26693cc4b8f
chunksPath: /data6/hdds/hdds/326a5fe1-e63c-44b6-a57e-2f858fe4eaa7/current/containerDir0/262574/chunks
containerDBType: RocksDB
containerID: 262574
containerType: KeyValueContainer
layOutVersion: 2
maxSize: 5368709120
metadata: {}
metadataPath: /data6/hdds/hdds/326a5fe1-e63c-44b6-a57e-2f858fe4eaa7/current/containerDir0/262574/metadata
originNodeId: 26395209-1233-439a-9020-0ad8d6a8248e
originPipelineId: cf7bd510-ff60-4d35-9746-7463d30e13af
state: CLOSED

 

A typical SCM group consists of 3 SCMs. If each of the them has its own scmUuid and clusterID, it will cause chaos in Datanode side.

For now, we just hard code the clusterID and scmUuid on HDDS-2823, so that we can setup a SCM raft group.

We need figure out a decent solution when merging back HDDS-2823 back to master. 

Issue Links

• is fixed by: HDDS-5432 Enable downgrade testing after 1.1.0 release
• links to: GitHub Pull Request #1649