FengQingxin created IOTDB-2134:
----------------------------------
Summary: CLONE - Incorrect count result of nodes and devices in
cluster
Key: IOTDB-2134
URL: https://issues.apache.org/jira/browse/IOTDB-2134
Project: Apache IoTDB
Issue Type: Bug
Components: Core/Cluster
Reporter: FengQingxin
Assignee: Eric Pai
Fix For: master branch
This bug can be reappeared by replaying the test case
IoTDBMetadataFetchIT.showCountDevices and IoTDBMetadataFetchIT.showCountNodes.
!image-2021-11-17-15-33-09-569.png!
!image-2021-11-17-15-33-32-630.png!
!image-2021-11-17-15-33-49-732.png!
The root causes are:
1. In nodes count implemented in ClusterPlanExecutor, if there's a SG
{_}root.a.b{_}, and a query {_}COUNT NODE root LEVEL=1{_}, we should return 1
immediately instead of sending to the SG leader, which returns 0.
In a word, for any LEVEL value being less than a storage group path nodes
length, we can get the result without sending to other nodes to query.
2.In devices count implemented in ClusterPlanExecutor, the query path should be
set properly if it should be transmitted to other nodes.
e.g. If we have a SG _root.a.b.c_ and a query {_}COUNT DEVICE root.a.b{_}, we
should pass _root.a.b.c_ to other nodes as it's the actual storage group name,
not _root.a.b_
if we have a SG _root.a_ and a query {_}COUNT DEVICE root.a.b{_}, we should
pass _root.a.b_ to other nodes as it's the actual path we queried.
In a word, at least one SG should be the prefix path of the one passed to other
nodes.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
