Updated on 2023-05-15 GMT+08:00

MRS 2.1.0.10 Patch Description

Basic Information

Table 1 Basic information

Patch Version

MRS 2.1.0.10

Release Date

2020-09-21

Resolved Issues

List of resolved issues in MRS 2.1.0.10:

MRS Manager

New queue configurations in the capacity-schedule.xml file will not be lost during cluster scale-out after the patch is installed.

Full-link monitoring can be rolled back.

Big data components

Hive permission assignment failure on Spark is resolved.

If no queue is specified, tasks are submitted to the launcher-job queue by default. Task running will not be affected.

List of resolved issues in MRS 2.1.0.9:

MRS Manager

The MRS Executor memory overflow is resolved.

Optimized the cluster scale-out process.

The problem that the SQL statement is incorrectly combined when the value of SparkSQL contains spaces is resolved.

The problem that HiveSQL jobs fail to be submitted occasionally is resolved.

The permission control for downloading the keytab file is optimized.

Big data components

When the Presto role name contains uppercase letters, the permission model can take effect.

The problem that Hive partitions are deleted slowly is resolved.

The problem that the token expires after Spark runs for a long time is resolved.

List of resolved issues in MRS 2.1.0.8:

MRS Manager

The problem that the ECS API traffic is limited when OBS is accessed through an agency has been solved.

Multiple users can log in to MRS Manager at the same time.

Full-link monitoring is supported.

MRS big data components

Carbon 2.0 has been upgraded.

The HBASE-18484 issue has been solved.

List of the resolved issues in MRS 2.1.0.7:

MRS Manager

The problem that data and files are displayed incorrectly if a field contains a newline character in the DLF+Presto query has been solved.

The Presto query result can be saved as a file.

List of resolved issues in MRS 2.1.0.6:

MRS Manager

The problem that the disk I/O usage of monitoring data is inaccurate has been solved.

The problem that the Spark job status is not updated occasionally has been solved.

The problem that the job running failure has been solved.

The patch mechanism has been optimized.

MRS big data components

The HBase exceptions are rectified.

The problem that the system responds slowly when Hive roles are bound to permissions has been solved.

List of resolved issues in MRS 2.1.0.5:

MRS big data components

Impala supports the ObsFileSystem function.

The timeout period of the MRS Manager page and the native pages of components can be configured.

The Hive privilege binding freezing problem has been solved.

The data connection failure has been solved.

List of resolved issues in MRS 2.1.0.3:

MRS Manager

Problems of Manager executor's high concurrent job submission have been solved.

MRS big data components

Data insertion failure in hive on tez has been fixed.

List of the resolved issues in MRS 2.1.0.2:

MRS Manager

No monitoring information is displayed after NodeAgent is restarted.

When a job is under submission for a long time, memory overflow occurs in the manager executor process.

Job submission is supported. manager executor can be used to configure high concurrency.

New Kafka topics are not displayed on the MRS Manager management plane.

When you call security cluster's APIs to submit the Spark Submit job and perform operations on an HBase table, the permission control on the HBase table does not take effect.

The MRS Manager patch mechanism has been optimized.

MRS big data components

The slow running of the load data inpath command executed by Spark has been optimized.

Column names containing the dollar sign ($) can be used in Spark table creation.

OBS-related problems have been solved.

List of the resolved issues in MRS 2.1.0.1:

MRS Manager

The return results of Hive SQL statements submitted by V2 jobs have been optimized, and the issue that V2 jobs fail to be submitted using an agency token has been solved.

MRS big data components

HiveServer out of memory (OOM) has been solved for MRS Hive: HIVE-10970 and HIVE-22275.

Compatibility with Other Patches

The MRS 2.1.0.10 patch package contains all patches released for MRS 2.1.0.

Vulnerability Disclosure

The remote code execution vulnerability of the Spark has been fixed. For details about the vulnerability, see CVE-2020-9480

Impact of Patch Installation

  • During the installation of the MRS 2.1.0.10 patch, MRS Manager will be restarted, and the components such as Hive, Impala, Spark, HDFS, Yarn, MapReduce, Presto, HBase, Tez, and related dependent services will be restarted in rolling mode. During the restart of MRS Manager, services are temporarily unavailable but services are not interrupted during the rolling restart.
  • After installing the MRS 2.1.0.10 patch, you need to download and install all clients again, including the original clients of Master nodes and the clients used by other nodes of VPC (that is, the clients that you set up).
    • You are advised to back up the old clients before reinstalling the new ones.
    • If you have modified client configurations based on the service scenario, modify them again after reinstalling the clients.
  • (Optional) The timeout interval of the MRS Manager page and the native page of the component can be configured. You need to manually modify the following configuration:
    1. Change the session timeout interval of the web and CAS services on all Master nodes.
      1. Change the value of <session-timeout>20</session-timeout> in /opt/Bigdata/tomcat/webapps/cas/WEB-INF/web.xml. The unit is minute.
      2. Change the value of <session-timeout>20</session-timeout> in /opt/Bigdata/tomcat/webapps/web/WEB-INF/web.xml. The unit is minute.
    2. Change the TGT validity period of the CAS on all Master nodes.

      Change 1200 in p:maxTimeToLiveInSeconds="${tgt.maxTimeToLiveInSeconds:1200} and p:timeToKillInSeconds="${tgt.timeToKillInSeconds:1200}" in /opt/Bigdata/tomcat/webapps/cas/WEB-INF/spring-configuration/ticketExpirationPolicies.xml to the corresponding timeout interval, in seconds.

    3. Restart the Tomcat service on the active Master node.
      1. On the active Master node, run the netstat -anp |grep 28443 |grep LISTEN command as user omm to query the Tomcat process ID.
      2. Run the kill -9 {pid} command, in which {pid} indicates the process ID obtained in the previous step.
      3. Wait for the process to automatically restart. You can run the netstat -anp |grep 28443 |grep LISTEN command to check whether the process is started. If the command output is displayed, the process is started successfully.
    4. Add or modify configuration items for each component. The values of the configuration items are the same as the timeout interval, in seconds.
      • HDFS/MapReduce/Yarn: Add the customized configuration item http.server.session.timeout.secs.
      • Spark: Change the value of spark.session.maxAge.
      • Hive: Add the customized configuration item http.server.session.timeout.secs.

      When saving the configuration items, you can choose not to restart the affected services or instances. Restart the services or instances when the service is not busy.