[
https://issues.apache.org/jira/browse/PHOENIX-6907?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17738304#comment-17738304
]
ASF GitHub Bot commented on PHOENIX-6907:
-----------------------------------------
virajjasani commented on code in PR #1598:
URL: https://github.com/apache/phoenix/pull/1598#discussion_r1245787461
##########
phoenix-core/src/main/java/org/apache/phoenix/iterate/ExplainTable.java:
##########
@@ -284,6 +367,102 @@ protected void explain(String prefix, List<String>
planSteps,
}
}
+ /**
+ * Retrieve region locations and set the values in the explain plan output.
+ *
+ * @param planSteps list of plan steps to add explain plan output to.
+ * @param explainPlanAttributesBuilder explain plan v2 attributes builder
instance.
+ * @param scansList list of the list of scans, to be used for parallel
scans.
+ */
+ private void getRegionLocations(List<String> planSteps,
+ ExplainPlanAttributesBuilder
explainPlanAttributesBuilder,
+ List<List<Scan>> scansList) {
+ Future<String> task = executorService.submit(
+ () ->
getRegionLocationsForExplainPlan(explainPlanAttributesBuilder, scansList));
+ try {
+ String regionLocations = task.get(5, TimeUnit.SECONDS);
+ planSteps.add(regionLocations);
+ } catch (InterruptedException | ExecutionException | TimeoutException
e) {
+ LOGGER.error("Unable to retrieve region locations task result in
5s", e);
+ task.cancel(true);
+ }
Review Comment:
this would be our own timeout, and if we can't get region locations on time,
we rather cancel the task than letting hbase client retry for very long time.
@jpisaac does this sound good to you?
> Explain Plan should output region locations with servers
> --------------------------------------------------------
>
> Key: PHOENIX-6907
> URL: https://issues.apache.org/jira/browse/PHOENIX-6907
> Project: Phoenix
> Issue Type: Improvement
> Reporter: Viraj Jasani
> Assignee: Viraj Jasani
> Priority: Major
> Fix For: 5.2.0, 5.1.4
>
>
> As of today, explain plan provides details related to scan key range size,
> size of the scan lists (for parallel scans), hints, client/server sorted by,
> order by, server filter etc. It provides as much close details to the
> physical plans as possible.
> If we could also make explain plan output to include region locations that
> also contains regionserver details on which the region is currently residing,
> this might turn out to be really good insights. For full table scans, this
> also means that we would end up including very higher number of region
> locations and it would not be ideal to print all of them in the output. We
> can enforce a limit on the num of region locations to print (default value
> could be 10 or so). For point lookup and range scans, we can find the region
> locations with regionserver as quite useful for the real time analysis of
> slowness experienced by the given query.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
