[
https://issues.apache.org/jira/browse/HIVE-25842?focusedWorklogId=710722&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-710722
]
ASF GitHub Bot logged work on HIVE-25842:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 18/Jan/22 18:18
Start Date: 18/Jan/22 18:18
Worklog Time Spent: 10m
Work Description: klcopp commented on a change in pull request #2916:
URL: https://github.com/apache/hive/pull/2916#discussion_r787032118
##########
File path:
ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/metrics/DeltaFilesMetricReporter.java
##########
@@ -512,7 +284,177 @@ private void shutdown() {
}
}
- public static class DeltaFilesMetadata implements Serializable {
- public String dbName, tableName, partitionName;
+ public static void updateMetricsFromInitiator(AcidDirectory dir, String
dbName, String tableName, String partitionName,
+ Configuration conf, TxnStore txnHandler) {
+ LOG.debug("Updating delta file metrics from initiator");
+ double deltaPctThreshold = MetastoreConf.getDoubleVar(conf,
MetastoreConf.ConfVars.METASTORE_DELTAMETRICS_DELTA_PCT_THRESHOLD);
+ int deltasThreshold = MetastoreConf.getIntVar(conf,
MetastoreConf.ConfVars.METASTORE_DELTAMETRICS_DELTA_NUM_THRESHOLD);
+ int obsoleteDeltasThreshold = MetastoreConf.getIntVar(conf,
+
MetastoreConf.ConfVars.METASTORE_DELTAMETRICS_OBSOLETE_DELTA_NUM_THRESHOLD);
+ try {
+ // We have an AcidDir from the initiator, therefore we can use that to
calculate active,small, obsolete delta
+ // count
+ long baseSize = getBaseSize(dir);
+
+ int numDeltas = dir.getCurrentDirectories().size();
+ int numSmallDeltas = 0;
+
+ for (AcidUtils.ParsedDelta delta : dir.getCurrentDirectories()) {
+ long deltaSize = getDirSize(delta, dir.getFs());
+ if (baseSize != 0 && deltaSize / (float) baseSize < deltaPctThreshold)
{
+ numSmallDeltas++;
+ }
+ }
+
+ int numObsoleteDeltas = dir.getObsolete().size();
Review comment:
I'm seriously wondering if we should.
Cons:
- the metric name is "obsolete deltas" would == obsolete + aborted deltas
- We already have metrics about the amount of aborts in the system (gotten
from metadata)
Pros:
- The Cleaner's job is to remove aborted directories as well ; so including
metrics about aborted directories would help with observability of Cleaner
health
- Aborted directories can clog up the file system just as much as obsolete
directories
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: gitbox-unsubscr...@hive.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
Issue Time Tracking
-------------------
Worklog Id: (was: 710722)
Time Spent: 3h 20m (was: 3h 10m)
> Reimplement delta file metric collection
> ----------------------------------------
>
> Key: HIVE-25842
> URL: https://issues.apache.org/jira/browse/HIVE-25842
> Project: Hive
> Issue Type: Improvement
> Reporter: László Pintér
> Assignee: László Pintér
> Priority: Major
> Labels: pull-request-available
> Time Spent: 3h 20m
> Remaining Estimate: 0h
>
> FUNCTIONALITY: Metrics are collected only when a Tez query runs a table
> (select * and select count( * ) don't update the metrics)
> Metrics aren't updated after compaction or cleaning after compaction, so
> users will probably see "issues" with compaction (like many active or
> obsolete or small deltas) that don't exist.
> RISK: Metrics are collected during queries – we tried to put a try-catch
> around each method in DeltaFilesMetricsReporter but of course this isn't
> foolproof. This is a HUGE performance and functionality liability. Tests
> caught some issues, but our tests aren't perfect.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
