rongchuan.jin created SPARK-34800:
-------------------------------------
Summary: Use fine-grained lock in SessionCatalog.tableExists
Key: SPARK-34800
URL: https://issues.apache.org/jira/browse/SPARK-34800
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 3.1.1
Reporter: rongchuan.jin
We have modified the underlying hive meta store which a different hive >
database is placed in its own shard for performance. However, we found that >
the synchronized limits the concurrency, we would like to fix it.
Related jstack trace like following:
{code:java}
"http-nio-7070-exec-257" #19961734 daemon prio=5 os_prio=0
tid=0x00007f45f4ce1000 nid=0x1a85e6 waiting for monitor entry
[0x00007f45949df000]
java.lang.Thread.State: BLOCKED (on object monitor)
at
org.apache.spark.sql.catalyst.catalog.SessionCatalog.tableExists(SessionCatalog.scala:415)
- waiting to lock <0x000000011d983d90> (a
org.apache.spark.sql.catalyst.catalog.SessionCatalog)
at
org.apache.spark.sql.catalyst.catalog.SessionCatalog.requireTableExists(SessionCatalog.scala:185)
at
org.apache.spark.sql.catalyst.catalog.SessionCatalog.getTableMetadata(SessionCatalog.scala:430)
at org.apache.spark.sql.DdlOperation$.getTableDesc(SourceFile:123)
at org.apache.spark.sql.DdlOperation.getTableDesc(SourceFile)
...
{code}
we fixed as discussed in mail
[http://mail-archives.apache.org/mod_mbox/spark-dev/202103.mbox/browser]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
