Trino / Presto Integration
Monitor query execution metrics, worker node health, and connector performance across your Trino and Presto clusters. Get per-query resource tracking and AI root cause analysis before slow queries impact your users.
How It Works
Install the TigerOps Trino Agent
Deploy the TigerOps agent on your Trino coordinator. It connects to the Trino REST API and JMX endpoint to discover all worker nodes, catalogs, and active queries automatically.
Enable JMX Metrics Export
Add the TigerOps JMX exporter as a Trino plugin in your jvm.config. All coordinator and worker MBeans — query queues, memory pools, task counts, and split scheduling — are collected out of the box.
Configure Per-Catalog Alerts
Set connector-specific alert thresholds for Hive, Iceberg, Delta Lake, and JDBC catalogs. TigerOps tracks connector worker utilization, failed splits, and catalog metadata fetch latency independently.
Map Queries to Application Traces
TigerOps links slow Trino queries to application-level traces via query ID propagation. You can see exactly which API endpoint triggered a slow SQL query and what the downstream impact was.
What You Get Out of the Box
Query Execution Metrics
Active, queued, and failed query counts with execution time histograms, CPU time per query, input data scanned, and output rows produced per coordinator and cluster.
Worker Node Health
Per-worker active task counts, split processing rates, heap memory usage, GC pause times, and network I/O rates. Detect stragglers and uneven work distribution instantly.
Connector Performance
Per-catalog connector request latency, split enumeration times, metadata cache hit rates, and failed connector calls for Hive, Iceberg, Delta Lake, and all JDBC catalogs.
Memory Pool Monitoring
Coordinator general and reserved memory pool utilization, per-query memory allocation, OOM killer trigger rates, and query memory revocation events for capacity planning.
Task & Stage Scheduling
Task scheduling latency, stage scheduling queue depth, split assignment rates, and pipeline driver execution stall counts to identify scheduling bottlenecks across large queries.
AI Slow Query Root Cause
When query p99 latency spikes, TigerOps AI pinpoints whether the cause is a straggler worker, a hot connector, a memory spill event, or a skewed partition — with suggested resolution steps.
TigerOps Agent for Trino
Configure the TigerOps agent to collect Trino JMX and REST API metrics and forward them to your TigerOps workspace.
# TigerOps Trino / Presto Agent Configuration
# Install: curl -sSL https://install.atatus.net/agent | sh
receivers:
trino_jmx:
jmx_url: "service:jmx:rmi:///jndi/rmi://trino-coordinator:9080/jmxrmi"
collection_interval: 15s
mbeans:
- "trino.execution:name=QueryManager"
- "trino.execution:name=TaskManager"
- "trino.memory:name=ClusterMemoryManager"
- "trino.failuredetector:name=HeartbeatFailureDetector"
trino_rest_api:
coordinator_url: "http://trino-coordinator:8080"
# Poll active queries for per-query resource tracking
query_poll_interval: 10s
# Poll worker list for node health
worker_poll_interval: 30s
auth:
type: basic
username: tigerops
password_env: TRINO_PASSWORD
exporters:
tigerops:
endpoint: "https://ingest.atatus.net/api/v1/write"
bearer_token: "${TIGEROPS_API_KEY}"
send_interval: 15s
# Alert thresholds
alerts:
query_failure_rate_pct: 5
active_query_queue_depth: 50
worker_heap_usage_pct: 90
coordinator_memory_pool_pct: 85
query_execution_time_p99_s: 300Common Questions
Does TigerOps support both Trino and Presto?
Yes. TigerOps supports Trino (formerly PrestoSQL) 350+ and PrestoDB 0.270+. The agent auto-detects which distribution is running via the version API and applies the correct metric namespace mapping for each. Ahana, Starburst, and AWS Athena v3 are also supported.
How does TigerOps track per-query resource consumption?
TigerOps polls the Trino /v1/query endpoint on the coordinator to collect per-query CPU time, memory allocation, input bytes, and execution state. Long-running queries are tracked in real time and alert when they exceed your configured resource thresholds.
Can I monitor Iceberg and Delta Lake table scan performance?
Yes. TigerOps tracks the Iceberg and Delta Lake catalog connector metrics including manifest file read latency, partition pruning effectiveness, and snapshot resolution time. You can correlate slow table scans with file count growth over time.
How does TigerOps handle Trino cluster autoscaling events?
TigerOps tracks worker join and leave events from the coordinator node list. When workers scale in or out, TigerOps annotates the metric timeline so you can correlate performance changes with cluster size changes automatically.
Can TigerOps alert on resource group queue depth?
Yes. TigerOps reads Trino resource group metrics via the /v1/resourceGroupInfo endpoint and alerts when a group queue exceeds your threshold. This lets you detect workload saturation before queries start failing with resource exhaustion errors.
Stop Discovering Slow Trino Queries After Users Complain
Per-query resource tracking, worker health monitoring, and AI root cause analysis. Deploy in 5 minutes.