HDDS-5022. SCM get roles command should provide Ratis Leader/Follower… #2094
Conversation
avijayanhwx
left a comment
avijayanhwx
left a comment
There was a problem hiding this comment.
Thanks for working on this @ghuangups. Left a comment inline.
There was a problem hiding this comment.
I am not sure if we should rely on the getAdminAddress() call to be null to determine if this node is LEADER or FOLLOWER. Can we do something similar to OM? The SCM that is returning the answer will be leader (since read calls go only to Leader), and the other nodes will be followers.
There was a problem hiding this comment.
Also, can we use org.apache.ratis.proto.RaftProtos.RaftPeerRole#LEADER/FOLLOWER instead of hardcoded strings?
There was a problem hiding this comment.
private static RaftPeer.Builder raftPeerBuilderFor(DatanodeDetails dn) {
return RaftPeer.newBuilder()
.setId(toRaftPeerId(dn))
.setAddress(toRaftPeerAddress(dn, Port.Name.RATIS_SERVER))
.setAdminAddress(toRaftPeerAddress(dn, Port.Name.RATIS_ADMIN))
.setClientAddress(toRaftPeerAddress(dn, Port.Name.RATIS));
}
I consider that each raft peer has its own admin address, no matter if it is leader or follower.
Better use NetUtils.isLocalAddress() to determine the peer whose address equals local, and assign him the leader role. Since SCM will do leader check before handle this request, the local SCM will be the leader.
|
Somehow the repo rejected my push. I had to use -f to overwrite. Created a new PR: #2098. |
3 participants
… information.
What changes were proposed in this pull request?
"ozone admin scm roles" command currently returns scm role info like this: [scm1:9865, scm2:9865, scm3:9865]. This change is to add: LEADER and FOLLOWER info in the return. The new result from the command is now look like this: [scm3:9865:FOLLOWER, scm2:9865:FOLLOWER, scm1:9865:LEADER]
What is the link to the Apache JIRA
(Please create an issue in ASF JIRA before opening a pull request,
and you need to set the title of the pull request which starts with
the corresponding JIRA issue number. (e.g. HDDS-XXXX. Fix a typo in YYY.)
https://issues.apache.org/jira/browse/HDDS-5022
How was this patch tested?
Manually tested locally by creating a SCM HA cluster with 3 SCM nodes and executed command: ozone admin scm roles from one of the SCM node. With this change, the command result was: [scm3:9865:FOLLOWER, scm2:9865:FOLLOWER, scm1:9865:LEADER].