[SLIDER-1169] Slider not honoring zookeeper quorum values passed

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: Slider 0.91
• Fix Version/s: Slider 0.92
• Component/s: appmaster

Environment

RHEL-6 (64 Bit)

• Assignee: Daniel Murnane
• Reporter: Sandeep Nemuri
• Created: 26/Sep/16 06:47
• Updated: 06/Mar/17 21:23
• Resolved: 26/Sep/16 23:39

Description

PROBLEM :

Customer has created a Slider App by passing zookeeper quorum using below command :

slider create test --template appConfig.json --resources resources.json --zkhosts sandy234new1.hwxblr.com:2181,sandy234new3.hwxblr.com:2181,sandy234new2.hwxblr.com:2181

Below is the application log, which show us that it only picks the 1st zookeeper.

2016-09-15 15:44:29,052 [main] INFO  appmaster.SliderAppMaster - Loading slider-server.xml at file:/hadoop/yarn/local/usercache/root/appcache/application_1473930641993_0005/container_e04_1473930641993_0005_01_000001/confdir/slider-server.xml
2016-09-15 15:44:29,077 [main] INFO  appmaster.SliderAppMaster - AM configuration:
hadoop.registry.zk.quorum=sandy234new1.hwxblr.com:2181
hadoop.registry.zk.root=/registry
yarn.resourcemanager.scheduler.address=0.0.0.0:8030

BUSINESS IMPACT : Slider throws exceptions when 1st zookeeper goes down (Since it only picks 1st zookeeper) and this is impacting the AM.

STEPS TO REPRODUCE:

Launch a Hbase app using step 1 & 2.
1) slider create test --template appConfig.json --resources resources.json --zkhosts sandy234new1.hwxblr.com:2181,sandy234new3.hwxblr.com:2181,sandy234new2.hwxblr.com:2181

This will launch a application in RM.

From the RM UI --> application -> logs

first line will be as below :

2016-09-15 15:44:29,052 [main] INFO  appmaster.SliderAppMaster - Loading slider-server.xml at file:/hadoop/yarn/local/usercache/root/appcache/application_1473930641993_0005/container_e04_1473930641993_0005_01_000001/confdir/slider-server.xml
2016-09-15 15:44:29,077 [main] INFO  appmaster.SliderAppMaster - AM configuration:
hadoop.registry.zk.quorum=sandy234new1.hwxblr.com:2181
hadoop.registry.zk.root=/registry
yarn.resourcemanager.scheduler.address=0.0.0.0:8030