[
https://issues.apache.org/jira/browse/IGNITE-8189?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16527586#comment-16527586
]
Amelchev Nikita commented on IGNITE-8189:
-----------------------------------------
[~sergey-chugunov], I have suggestions:
- make *deleteAll* method takes only a collection with full paths, without
parent directory parameter.
- add *list<String> makePaths(@Nullable String parent, List<String> paths)*
private method which returns the full paths for joining few requests with
different parents.
This will allow us to include requests into one batch in *deleteFutureData()*
and *cleanupPreviousClusterData()* methods. What do you think?
> Improve ZkDistributedCollectDataFuture#deleteFutureData implementation
> ----------------------------------------------------------------------
>
> Key: IGNITE-8189
> URL: https://issues.apache.org/jira/browse/IGNITE-8189
> Project: Ignite
> Issue Type: Improvement
> Components: zookeeper
> Reporter: Sergey Chugunov
> Assignee: Amelchev Nikita
> Priority: Major
>
> Three issues need to be improved in implementation:
> * two more deleteIfExists methods within the *deleteFutureData* to be
> included in batching *deleteAll* operation;
> * if request exceeds ZooKeeper max size limit fallback to one-by-one deletion
> should be used (related ticket IGNITE-8188);
> * ZookeeperClient#deleteAll implementation may throw NoNodeException is case
> of concurrent operation removing the same nodes, in this case fallback to
> one-by-one deletion should be used too.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
