Configuring Impala executor groups
Configure Impala executor groups to partition executors into separate logical groups, enabling workload isolation and resource management across different workload types.
Setting the executor group name by using the dedicated Impala Daemon Executor Group (impalad_executor_group)Cloudera Manager property. Cloudera Manager then sets the matching executor group sets on the coordinators.
Executor group names
Each executor group is set with a single value in the [***SET
NAME***]-[***SUFFIX***] format , such as
tier1-g1.
The following table describes the components of the executor group naming syntax:
| Term | Example | Description |
|---|---|---|
| Executor group name | tier1-g1 |
Specifies the entire value. Executors with the same value form one executor group. |
| Set name | tier1 |
Specifies the part before the hyphen. Groups with the same set name form an executor group set that serves one resource pool. |
| Suffix | g1 |
Specifies the identifier after the hyphen that distinguishes groups in the same set. |
When configuring executor groups, adhere to the following requirements:
- Character restrictions: Use only letters, digits, and underscores, with exactly one hyphen between the set name and the suffix. Cloudera Manager rejects other values.
- Resource pool matching: The set name must match a resource pool. Create the pool
root.***SET NAME***-***SUFFIX***in Impala Admission Control. Cloudera Manager issues a warning if the pool is missing. - Concurrency and query routing: Add groups to a set to increase query concurrency. Each query runs entirely on one group of its set. Adding more groups to a set allows the set to run more queries concurrently.
- Consistent executor sizing: Assign every group in a set the same number of executors. In addition, ensure that all executors within a group set have identical resource allocations, such as memory, CPU, and network capacity. Impala plans queries by using a single group size and uniform executor sizing per set.
- Distinct set names: Do not use a set name that starts with another set name, such as
tier1andtier10. Impala matches groups to sets by prefix, sotier10-g1is also counted as part oftier1. - Role assignment: Leave the property value empty on coordinators. Executors with no value join the built-in default group.
For example, tier1-g1, tier1-g2 (configured with 5
executors each) and etl-g1 (configured with 10 executors) create two sets.
The tier1 set contains two groups and serves root.tier1;
the etl set contains one group and serves root.etl.
The executor group is now configured. Impala uses this group for admission control and cluster health validation. A group is healthy once at least one of its executors is running.
