Dynamically Change the HDInsight Yarn Capacity Allocation
Requirement:
The fundamental unit of YARN is a queue. The user
can submit a job to a specific queue. Each queue has a capacity
defined by cluster admin and accordingly share of resources are allocated to
the queue.
Requirement here is to re-define the YARN memory allocation using
scheduled job
Solution:
1. Identify the config type to update and retrieve
the latest yarn version. When any config type needs to update the whole
property set needs to be updated. So, copy the entire property set from existing
values.
2. Update the TAG with unique value for new
version. As a best practice use current timestamp as a TAG version
3. Edit the properties as needed and then update
the config type
4. The new version of the config type must be
added and applied to the cluster
We can use REST
APIs or a wrapper script (/var/lib/ambari-server/resources/scripts/configs.py
or /var/lib/ambari-server/resources/scripts/configs.sh) to edit configurations.
Implementation:
Step 1:
Go to this URL https://sparktest.azurehdinsight.net/api/v1/clusters/sparktest
and identify the latest YARN Version. Find the screenshot below
YARN
Queue Manager:
Cluster Config URL:
Step 2: Update
the property mentioned above as required
Step 3: Using
REST API we can submit the changes
curl -u "admin:pass" -H "X-Requested-By: ambari" -X PUT "http://hn0-sparketest:8080/api/v1/clusters/sparktestdev" -d '[{"Clusters":{
"desired_config":[{
"type":"capacity-scheduler",
"tag":"version1573054562219",
"properties":{
"yarn.scheduler.capacity.maximum-am-resource-percent" : "0.33",
"yarn.scheduler.capacity.maximum-applications" : "10000",
"yarn.scheduler.capacity.node-locality-delay" : "0",
"yarn.scheduler.capacity.queue-mappings-override.enable" : "false",
"yarn.scheduler.capacity.resource-calculator" : "org.apache.hadoop.yarn.util.resource.DefaultResourceCalculator",
"yarn.scheduler.capacity.root.accessible-node-labels" : "*",
"yarn.scheduler.capacity.root.acl_administer_queue" : "*",
"yarn.scheduler.capacity.root.capacity" : "100",
"yarn.scheduler.capacity.root.default.acl_submit_applications" : "*",
"yarn.scheduler.capacity.root.default.capacity" : "50",
"yarn.scheduler.capacity.root.default.maximum-capacity" : "80",
"yarn.scheduler.capacity.root.default.priority" : "0",
"yarn.scheduler.capacity.root.default.state" : "RUNNING",
"yarn.scheduler.capacity.root.default.user-limit-factor" : "10",
"yarn.scheduler.capacity.root.priority" : "0",
"yarn.scheduler.capacity.root.quality_analytics.acl_administer_queue" : "*",
"yarn.scheduler.capacity.root.quality_analytics.acl_submit_applications" : "*",
"yarn.scheduler.capacity.root.quality_analytics.capacity" : "40",
"yarn.scheduler.capacity.root.quality_analytics.maximum-capacity" : "100",
"yarn.scheduler.capacity.root.quality_analytics.minimum-user-limit-percent" : "100",
"yarn.scheduler.capacity.root.quality_analytics.ordering-policy" : "fair",
"yarn.scheduler.capacity.root.quality_analytics.ordering-policy.fair.enable-size-based-weight" : "false",
"yarn.scheduler.capacity.root.quality_analytics.priority" : "0",
"yarn.scheduler.capacity.root.quality_analytics.state" : "RUNNING",
"yarn.scheduler.capacity.root.quality_analytics.user-limit-factor" : "1",
"yarn.scheduler.capacity.root.queues" : "default,quality_analytics,thriftsvr",
"yarn.scheduler.capacity.root.thriftsvr.capacity" : "10",
"yarn.scheduler.capacity.root.thriftsvr.maximum-capacity" : "40",
"yarn.scheduler.capacity.root.thriftsvr.priority" : "0",
"yarn.scheduler.capacity.root.thriftsvr.user-limit-factor" : "10"
},
"service_config_version_note":"New config
version"}]}}]'
