You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@ambari.apache.org by Aravindan Vijayan <av...@hortonworks.com> on 2017/02/16 21:10:31 UTC
Review Request 56765: AMBARI-20056 : FIFO compaction resulting in too
many store files on large cluster
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56765/
-----------------------------------------------------------
Review request for Ambari, Dmytro Sen, Sumit Mohanty, and Sid Wagle.
Bugs: AMBARI-20056
https://issues.apache.org/jira/browse/AMBARI-20056
Repository: ambari
Description
-------
PROBLEM
On a 500 node cluster AMS can become in-operational after creating 18K store files and no compaction.
FIX
Set Data Tiered compaction for all tables except METRIC_RECORD. This will allow number of store files to go down over time.
As part of this effort, I refined the segment of the code where AMS HBase tables policies are set such that any compaction policy can be set for precision and aggregate tables.
Defaults
Precision Table (METRIC_RECORD)
hbase.hstore.defaultengine.compactionpolicy.class : org.apache.hadoop.hbase.regionserver.compactions.FIFOCompactionPolicy
Aggregate Tables
hbase.hstore.engine.class : org.apache.hadoop.hbase.regionserver.DateTieredStoreEngine
Diffs
-----
ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java b249d05
ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java ad05025
ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java 7bbd5aa
ambari-metrics/ambari-metrics-timelineservice/src/test/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/ITPhoenixHBaseAccessor.java 3bc1929
ambari-server/src/main/java/org/apache/ambari/server/upgrade/UpgradeCatalog250.java 0d0b8a2
ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml 76f8660
ambari-server/src/test/java/org/apache/ambari/server/upgrade/UpgradeCatalog250Test.java 112f212
Diff: https://reviews.apache.org/r/56765/diff/
Testing
-------
Manually tested.
mvn clean test on Ambari metrics.
Thanks,
Aravindan Vijayan
Re: Review Request 56765: AMBARI-20056 : FIFO compaction resulting in
too many store files on large cluster
Posted by Sid Wagle <sw...@hortonworks.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56765/#review165882
-----------------------------------------------------------
Fix it, then Ship it!
ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java (line 635)
<https://reviews.apache.org/r/56765/#comment237715>
We shuld make blocking store files configurable.
- Sid Wagle
On Feb. 16, 2017, 9:10 p.m., Aravindan Vijayan wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/56765/
> -----------------------------------------------------------
>
> (Updated Feb. 16, 2017, 9:10 p.m.)
>
>
> Review request for Ambari, Dmytro Sen, Sumit Mohanty, and Sid Wagle.
>
>
> Bugs: AMBARI-20056
> https://issues.apache.org/jira/browse/AMBARI-20056
>
>
> Repository: ambari
>
>
> Description
> -------
>
> PROBLEM
> On a 500 node cluster AMS can become in-operational after creating 18K store files and no compaction.
>
>
> FIX
> Set Data Tiered compaction for all tables except METRIC_RECORD. This will allow number of store files to go down over time.
>
> As part of this effort, I refined the segment of the code where AMS HBase tables policies are set such that any compaction policy can be set for precision and aggregate tables.
>
> Defaults
> Precision Table (METRIC_RECORD)
> hbase.hstore.defaultengine.compactionpolicy.class : org.apache.hadoop.hbase.regionserver.compactions.FIFOCompactionPolicy
>
> Aggregate Tables
> hbase.hstore.engine.class : org.apache.hadoop.hbase.regionserver.DateTieredStoreEngine
>
>
> Diffs
> -----
>
> ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java b249d05
> ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java ad05025
> ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java 7bbd5aa
> ambari-metrics/ambari-metrics-timelineservice/src/test/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/ITPhoenixHBaseAccessor.java 3bc1929
> ambari-server/src/main/java/org/apache/ambari/server/upgrade/UpgradeCatalog250.java 0d0b8a2
> ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml 76f8660
> ambari-server/src/test/java/org/apache/ambari/server/upgrade/UpgradeCatalog250Test.java 112f212
>
> Diff: https://reviews.apache.org/r/56765/diff/
>
>
> Testing
> -------
>
> Manually tested.
> mvn clean test on Ambari metrics.
>
>
> Thanks,
>
> Aravindan Vijayan
>
>
Re: Review Request 56765: AMBARI-20056 : FIFO compaction resulting in
too many store files on large cluster
Posted by Sumit Mohanty <sm...@hortonworks.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56765/#review166109
-----------------------------------------------------------
Ship it!
Ship It!
- Sumit Mohanty
On Feb. 20, 2017, 10:28 p.m., Aravindan Vijayan wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/56765/
> -----------------------------------------------------------
>
> (Updated Feb. 20, 2017, 10:28 p.m.)
>
>
> Review request for Ambari, Dmytro Sen, Sumit Mohanty, and Sid Wagle.
>
>
> Bugs: AMBARI-20056
> https://issues.apache.org/jira/browse/AMBARI-20056
>
>
> Repository: ambari
>
>
> Description
> -------
>
> PROBLEM
> On a 500 node cluster AMS can become in-operational after creating 18K store files and no compaction.
>
>
> FIX
> Set Data Tiered compaction for all tables except METRIC_RECORD. This will allow number of store files to go down over time.
>
> As part of this effort, I refined the segment of the code where AMS HBase tables policies are set such that any compaction policy can be set for precision and aggregate tables.
>
> Defaults
> Precision Table (METRIC_RECORD)
> hbase.hstore.defaultengine.compactionpolicy.class : org.apache.hadoop.hbase.regionserver.compactions.FIFOCompactionPolicy
>
> Aggregate Tables
> hbase.hstore.engine.class : org.apache.hadoop.hbase.regionserver.DateTieredStoreEngine
>
>
> Diffs
> -----
>
> ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java b249d05
> ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java ad05025
> ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java 7bbd5aa
> ambari-metrics/ambari-metrics-timelineservice/src/test/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/ITPhoenixHBaseAccessor.java 3bc1929
> ambari-server/src/main/java/org/apache/ambari/server/upgrade/UpgradeCatalog250.java 6a3e751
> ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml 76f8660
> ambari-server/src/test/java/org/apache/ambari/server/upgrade/UpgradeCatalog250Test.java 9e36e20
>
> Diff: https://reviews.apache.org/r/56765/diff/
>
>
> Testing
> -------
>
> Manually tested.
> mvn clean test on Ambari metrics.
>
>
> Thanks,
>
> Aravindan Vijayan
>
>
Re: Review Request 56765: AMBARI-20056 : FIFO compaction resulting in
too many store files on large cluster
Posted by Aravindan Vijayan <av...@hortonworks.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/56765/
-----------------------------------------------------------
(Updated Feb. 20, 2017, 10:28 p.m.)
Review request for Ambari, Dmytro Sen, Sumit Mohanty, and Sid Wagle.
Bugs: AMBARI-20056
https://issues.apache.org/jira/browse/AMBARI-20056
Repository: ambari
Description
-------
PROBLEM
On a 500 node cluster AMS can become in-operational after creating 18K store files and no compaction.
FIX
Set Data Tiered compaction for all tables except METRIC_RECORD. This will allow number of store files to go down over time.
As part of this effort, I refined the segment of the code where AMS HBase tables policies are set such that any compaction policy can be set for precision and aggregate tables.
Defaults
Precision Table (METRIC_RECORD)
hbase.hstore.defaultengine.compactionpolicy.class : org.apache.hadoop.hbase.regionserver.compactions.FIFOCompactionPolicy
Aggregate Tables
hbase.hstore.engine.class : org.apache.hadoop.hbase.regionserver.DateTieredStoreEngine
Diffs (updated)
-----
ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java b249d05
ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java ad05025
ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java 7bbd5aa
ambari-metrics/ambari-metrics-timelineservice/src/test/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/ITPhoenixHBaseAccessor.java 3bc1929
ambari-server/src/main/java/org/apache/ambari/server/upgrade/UpgradeCatalog250.java 6a3e751
ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml 76f8660
ambari-server/src/test/java/org/apache/ambari/server/upgrade/UpgradeCatalog250Test.java 9e36e20
Diff: https://reviews.apache.org/r/56765/diff/
Testing
-------
Manually tested.
mvn clean test on Ambari metrics.
Thanks,
Aravindan Vijayan