Repository: carbondata Updated Branches: refs/heads/master 6fd778ab1 -> 44e70d08e
[CARBONDATA-2117]Fixed Syncronization issue in CarbonEnv Problem: When creating multiple session (100) session initialisation is failing with below error java.lang.IllegalArgumentException: requirement failed: Config entry enable.unsafe.sort already registered! Solution: Currently in CarbonEnv we are updating global configuration(shared) and location configuration in class level synchronized block. In case of multiple session class level lock will not work , need to add global level lock so only one thread will update the global configuration This closes #1908 Project: http://git-wip-us.apache.org/repos/asf/carbondata/repo Commit: http://git-wip-us.apache.org/repos/asf/carbondata/commit/44e70d08 Tree: http://git-wip-us.apache.org/repos/asf/carbondata/tree/44e70d08 Diff: http://git-wip-us.apache.org/repos/asf/carbondata/diff/44e70d08 Branch: refs/heads/master Commit: 44e70d08e0c73e2c65e9a0d147cbbbe965aaf9f7 Parents: 6fd778a Author: kumarvishal <kumarvishal.1...@gmail.com> Authored: Thu Feb 1 23:13:54 2018 +0530 Committer: Jacky Li <jacky.li...@qq.com> Committed: Sat Feb 3 17:36:40 2018 +0800 ---------------------------------------------------------------------- .../spark2/src/main/scala/org/apache/spark/sql/CarbonEnv.scala | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/carbondata/blob/44e70d08/integration/spark2/src/main/scala/org/apache/spark/sql/CarbonEnv.scala ---------------------------------------------------------------------- diff --git a/integration/spark2/src/main/scala/org/apache/spark/sql/CarbonEnv.scala b/integration/spark2/src/main/scala/org/apache/spark/sql/CarbonEnv.scala index 40035ce..6b12008 100644 --- a/integration/spark2/src/main/scala/org/apache/spark/sql/CarbonEnv.scala +++ b/integration/spark2/src/main/scala/org/apache/spark/sql/CarbonEnv.scala @@ -68,7 +68,8 @@ class CarbonEnv { // added for handling timeseries function like hour, minute, day , month , year sparkSession.udf.register("timeseries", new TimeSeriesFunction) - synchronized { + // acquiring global level lock so global configuration will be updated by only one thread + CarbonEnv.carbonEnvMap.synchronized { if (!initialized) { // update carbon session parameters , preserve thread parameters val currentThreadSesssionInfo = ThreadLocalSessionInfo.getCarbonSessionInfo
