You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues-all@impala.apache.org by "ASF subversion and git services (Jira)" <ji...@apache.org> on 2023/01/01 07:15:00 UTC
[jira] [Commented] (IMPALA-11812) Catalogd OOM due to lots of HMS FieldSchema instances
[ https://issues.apache.org/jira/browse/IMPALA-11812?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17653376#comment-17653376 ]
ASF subversion and git services commented on IMPALA-11812:
----------------------------------------------------------
Commit 77d80aeda653b3aecb8bc41bf867cc5a84ba1245 in impala's branch refs/heads/master from stiga-huang
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=77d80aeda ]
IMPALA-11812: Deduplicate column schema in hmsPartitions
A list of HMS Partitions will be created in many workloads in catalogd,
e.g. table loading, bulk altering partitions by ComputeStats or
AlterTableRecoverPartitions, etc. Currently, each of hmsPartition hold a
unique list of column schema, i.e. a List<FieldSchema>. This results in
lots of FieldSchema instances if the table is wide and lots of
partitions need to be loaded/operated. Though the strings of column
names and comments are interned, the FieldSchema objects could still
occupy the majority of the heap. See the histogram in JIRA description.
In reality, the hmsPartition instances of a table can share the
table-level column schema since Impala doesn't respect the partition
level schema.
This patch replaces column list in StorageDescriptor of hmsPartitions
with the table level column list to remove the duplications. Also add
some progress logs in batch HMS operations, and avoid misleading logs
when event-processor is disabled.
Tests:
- Ran exhaustive tests
- Add tests on wide table operations that hit OOM errors without this
fix.
Change-Id: I511ecca0ace8bea4c24a19a54fb0a75390e50c4d
Reviewed-on: http://gerrit.cloudera.org:8080/19391
Reviewed-by: Aman Sinha <am...@cloudera.com>
Tested-by: Impala Public Jenkins <im...@cloudera.com>
> Catalogd OOM due to lots of HMS FieldSchema instances
> -----------------------------------------------------
>
> Key: IMPALA-11812
> URL: https://issues.apache.org/jira/browse/IMPALA-11812
> Project: IMPALA
> Issue Type: Bug
> Components: Catalog
> Reporter: Quanlong Huang
> Assignee: Quanlong Huang
> Priority: Critical
> Attachments: MAT_dominator_tree.png, create_ext_tbl_with_5k_cols_50k_parts.sh, wide-table-loading-oom-FieldSchema-path2root.png, wide-table-loading-oom-histogram.png, wide-table-loading-oom-top-consumers.png
>
>
> For partitioned wide tables that have thousands of columns, catalogd might hit OOM in routines on them. E.g. when running AlterTableRecoverPartitions for all their partitions, or when initially loading all partitions of them.
> The direct reason is that the heap is full of HMS FieldSchema instances. Here is a histogram of the issue in a 4GB heap:
> {noformat}
> Class Name | Objects | Shallow Heap |
> --------------------------------------------------------------------------------------------
> org.apache.hadoop.hive.metastore.api.FieldSchema | 111,876,486 | 2,685,035,664 |
> java.lang.Object[] | 78,026 | 449,929,656 |
> char[] | 91,295 | 6,241,744 |
> java.util.ArrayList | 171,126 | 4,107,024 |
> java.util.HashMap | 71,135 | 3,414,480 |
> java.lang.String | 91,161 | 2,187,864 |
> java.util.concurrent.ConcurrentHashMap$Node | 59,614 | 1,907,648 |
> java.util.concurrent.atomic.LongAdder | 53,021 | 1,696,672 |
> org.apache.hadoop.hive.metastore.api.Partition | 22,374 | 1,610,928 |
> com.codahale.metrics.EWMA | 30,780 | 1,477,440 |
> com.codahale.metrics.LongAdderProxy$JdkProvider$1 | 53,021 | 1,272,504 |
> org.apache.hadoop.hive.metastore.api.StorageDescriptor | 22,376 | 1,253,056 |
> java.util.Hashtable$Entry | 36,921 | 1,181,472 |
> java.util.concurrent.atomic.AtomicLong | 39,444 | 946,656 |
> org.apache.hadoop.hive.metastore.api.SerDeInfo | 22,375 | 895,000 |
> byte[] | 1,686 | 668,480 |
> java.util.concurrent.ConcurrentHashMap$Node[] | 1,874 | 639,824 |
> com.codahale.metrics.ExponentiallyDecayingReservoir | 10,259 | 574,504 |
> java.util.HashMap$Node | 17,776 | 568,832 |
> org.apache.hadoop.hive.metastore.api.SkewedInfo | 22,375 | 537,000 |
> com.codahale.metrics.Meter | 10,260 | 492,480 |
> java.util.concurrent.ConcurrentSkipListMap | 10,260 | 492,480 |
> java.util.concurrent.locks.ReentrantReadWriteLock$NonfairSync| 10,259 | 492,432 |
> org.apache.impala.catalog.ColumnStats | 5,003 | 400,240 |
> Total: 24 of 6,158 entries; 6,130 more | 113,007,927 | 3,174,330,288 | {noformat}
> In the above case, these FieldSchema instances come from the list of hmsPartitions that is created locally by CatalogOpExecutor#alterTableRecoverPartitions(). The thread is 0x6d051abb8:
> !MAT_dominator_tree.png|width=825,height=308!
> Stacktrace:
> {noformat}
> Thread 0x6d051abb8
> at org.apache.hadoop.hive.metastore.api.StorageDescriptor.<init>(Lorg/apache/hadoop/hive/metastore/api/StorageDescriptor;)V (StorageDescriptor.java:216)
> at org.apache.hadoop.hive.metastore.api.StorageDescriptor.deepCopy()Lorg/apache/hadoop/hive/metastore/api/StorageDescriptor; (StorageDescriptor.java:256)
> at org.apache.impala.service.CatalogOpExecutor.createHmsPartitionFromValues(Ljava/util/List;Lorg/apache/hadoop/hive/metastore/api/Table;Lorg/apache/impala/analysis/TableName;Ljava/lang/String;)Lorg/apache/hadoop/hive/metastore/api/Partition; (CatalogOpExecutor.java:5787)
> at org.apache.impala.service.CatalogOpExecutor.alterTableRecoverPartitions(Lorg/apache/impala/catalog/Table;Ljava/lang/String;)V (CatalogOpExecutor.java:5678)
> at org.apache.impala.service.CatalogOpExecutor.alterTable(Lorg/apache/impala/thrift/TAlterTableParams;Ljava/lang/String;ZLorg/apache/impala/thrift/TDdlExecResponse;)V (CatalogOpExecutor.java:1208)
> at org.apache.impala.service.CatalogOpExecutor.execDdlRequest(Lorg/apache/impala/thrift/TDdlExecRequest;)Lorg/apache/impala/thrift/TDdlExecResponse; (CatalogOpExecutor.java:419)
> at org.apache.impala.service.JniCatalog.execDdl([B)[B (JniCatalog.java:260){noformat}
> *How this happen*
> When creating the list of hmsPartitions, we deep copy the StorageDescriptor which will also deep copy the column list:
> {code:java}
> alterTableRecoverPartitions()
> -> createHmsPartitionFromValues()
> -> StorageDescriptor sd = msTbl.getSd().deepCopy();{code}
> Impala doesn't respect the partition level schema (by design), we should share the list of FieldSchema across hmsPartitions.
> When loading partition metadata for such a table, we could also hit this issue. The HMS API "get_partitions_by_names" returns the list of hmsPartitions. Each of them reference a unique list of FieldSchemas. We should deduplicate them to share the same column list. FWIW, attached the heap analysis results for wide table loading OOM (4GB heap):
> !wide-table-loading-oom-histogram.png|width=623,height=695!
> !wide-table-loading-oom-top-consumers.png|width=1066,height=434!
> The FieldSchema instances come from the metadata loading thread:
> !wide-table-loading-oom-FieldSchema-path2root.png|width=1023,height=213!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscribe@impala.apache.org
For additional commands, e-mail: issues-all-help@impala.apache.org