Technologies/Trino/trino.task.running_tasks
TrinoTrinoMetric

trino.task.running_tasks

Tasks currently executing
Dimensions:None
Available on:PrometheusPrometheus (1)
Interface Metrics (1)
PrometheusPrometheus
Number of tasks currently executing
Dimensions:None
Sources
trino_running_taskstrino.io

Technical Annotations (28)

Configuration Parameters (11)
fault-tolerant-execution-standard-split-sizerecommended: 64MB
Standard split data size processed by tasks reading from source tables
fault-tolerant-execution-max-task-split-countrecommended: 2048
Maximum splits per task, protects against incorrect split weight
task.concurrencyrecommended: 4
Controls number of concurrent tasks per worker, set in reported configuration
task.max-drivers-per-taskrecommended: 8
Limits drivers per task, may affect parallelism when threads are blocked
task.http-timeout-threadsrecommended: 15-20
HTTP timeout thread pool size, exhaustion can cause worker communication issues
task.http-response-threadsrecommended: 300
HTTP response thread pool, insufficient threads can block worker communication
experimental.thread-per-driver-scheduler-enabledrecommended: false
enabled by default since v438, causes communication timeouts and prevents JMX metric access
query.remote-task.max-error-durationrecommended: 1m
Default timeout for remote task communication
task.min-drivers
default is task.max-worker-threads * 2; increase if resources underutilized
task.low-memory-killer.policyrecommended: least-waste
Avoids killing long-running tasks to prevent wasting completed work
retry-policyrecommended: TASK
Task-level killer only applies when retry-policy is TASK
Error Signatures (2)
SERVICE_UNAVAILABLEhttp status
java.util.concurrent.TimeoutException: Total timeout 10000 ms elapsedexception
CLI Commands (1)
EXPLAIN ANALYZE VERBOSEdiagnostic
Technical References (14)
worker parallelismconceptquery splitsconceptWAITINGconceptTIMED_WAITINGconceptio.trino.operator.ScanFilterAndProjectOperatorcomponentio.trino.parquet.reader.PrimitiveColumnReadercomponentstage 0conceptthread-per-driver schedulercomponenttrino.execution.executor:name=TaskExecutor.RunningSplitscomponenthttp://worker:9300/v1/jmx/mbeanfile pathcoordinatorcomponentremote taskcomponentbucketingconceptbloom filterconcept
Related Insights (8)
Improper task sizing prevents query completionwarning
▸
Worker parallelism drops to zero with threads stuck in WAITING statecritical
▸
Long-running tasks accumulate on single worker causing timeoutscritical
▸
Thread-per-driver scheduler causes worker communication timeouts and 503 errorscritical
▸
Remote task failures from coordinator communication timeoutwarning
▸
ETL workloads suffer I/O bottlenecks and uneven worker resource utilizationwarning
▸
Underutilized worker resources from low task.min-drivers settinginfo
▸
Tasks killed by low memory killer with inefficient waste handlingcritical
▸