Support Questions

Find answers, ask questions, and share your expertise

default queue accepting one mr job for hive while tez accepting more than one

avatar

default queue accepting one mr job for hive while tez accepting more than one.

default execution engine is Tez. in this case default queue running 3 different query parallely in 3 different session.

While i set hive.execution.engine=mr in this case i am able to open multiple hive session but default queue run only one query at a time.

My cluster is big and already tried multiple small job parallely for Tez. but setting hive.execution.engine=mr run only one job at a time in default queue.

1 ACCEPTED SOLUTION

avatar
Super Collaborator

Do you have the YARN Resource Manager screenshot when you run 3 mapred job?

http://<Active_RM_HOST>:8088/cluster/scheduler

From what i read from screen shots,

Maximum AM Resource 20% i.e. 20% of 391GB = 78GB.

value of: yarn.app.mapreduce.am.resource.mb and tez.am.resource.memory.mb will determine how many AMs can fit in to run concurrently.

View solution in original post

7 REPLIES 7

avatar
Super Collaborator

Hi @Alpesh Virani

There are several possibilities. When you have multiple hive sessions open with execution engine "mr", can you tell us:

1. How much resources you have used / available to run in "default" queue ?

Check this on YARN RM UI > Scheduler

2. If you have 100% available in "default" queue, check the "am" container size "yarn.app.mapreduce.am.resource.mb" and check the "Maximum AM Resource" for "default" queue,

see if queue has enough resources to run multiple "am" "mr" containers.

avatar

screen-shot-2018-03-14-at-83913-am.png

See th attached file it is inheriting all the resource and it is 100%.

avatar
Super Collaborator

Hi @Alpesh Virani

Can you also share the Resource Manager UI screenshot. This will tell what is the actual usage for your queue.

http://<Active_RM_HOST>:8088/cluster/scheduler

avatar

screen-shot-2018-03-14-at-84702-am.png

If you see i am able to run 3 TEZ job. but when i run 3 mapred job in hive.execution.engine=mr in hive. it is accepting one at a time despite enough space.

avatar
Super Collaborator

Do you have the YARN Resource Manager screenshot when you run 3 mapred job?

http://<Active_RM_HOST>:8088/cluster/scheduler

From what i read from screen shots,

Maximum AM Resource 20% i.e. 20% of 391GB = 78GB.

value of: yarn.app.mapreduce.am.resource.mb and tez.am.resource.memory.mb will determine how many AMs can fit in to run concurrently.

avatar

Thanks you very much for your prompt reply. I am able to run map reduce with 300 GB usage. I think i need to look some other parameter. your hint may help me. currently i can not run it but will do it soon.

avatar
Super Collaborator

Please accept an Answer, so that we can mark this request close.