HDDS-7402. Adapt CommandQueue to track the count of each queued command type#3891
HDDS-7402. Adapt CommandQueue to track the count of each queued command type#3891umamaheswararao merged 4 commits intoapache:masterfrom
Conversation
swamirishi
left a comment
swamirishi
left a comment
There was a problem hiding this comment.
Left a comment, for making the code more modular
| // This list is used as default return value. | ||
| private static final List<SCMCommand> DEFAULT_LIST = new ArrayList<>(); | ||
| private final Map<UUID, Commands> commandMap; | ||
| private final Map<UUID, Map<SCMCommandProto.Type, Integer>> summaryMap; |
There was a problem hiding this comment.
Would it be better if we add another class to handle Commands for one Datanode separately? Like something like DatanodeQueue. So summary and commands can go together in one place.
There was a problem hiding this comment.
I guess that is something we could do in the future if there were more things to track on a datanode by datanode basis. I'd like to avoid too much change here, and a class per DN like a DatanodeQueue would look fairly similar I think. Something like this could easily be added later if we need it, as it will be behind the CommandQueue interface, and the changes would just be internal to the class.
There was a problem hiding this comment.
Maybe the summary could go into the commands class, since we already have that
There was a problem hiding this comment.
I see we are doing the same map operation twice here
There was a problem hiding this comment.
Yea putting the summary into the commands class might make it tidier. I will try that tomorrow.
3 participants
What changes were proposed in this pull request?
Currently, all commands queued for a datanode are stored in an ArrayList, which is sent to the DN on each heartbeat.
In order to limit the number of commands queued on a DN at any time, it would be useful to be able to query the current queued count of each command type, allowing Replication Manager, for example, to stop sending more work to the DN until it has free capacity.
What is the link to the Apache JIRA
https://issues.apache.org/jira/browse/HDDS-7402
How was this patch tested?
New unit tests