Trino CrashLoopBackOff from a missing BigQuery connector JVM argument

Entities identified

Hypotheses tree

Ruled-out branches are collapsed. Open any one to read the check that eliminated it.

Image: trinodb/trino:475 (pulled successfully)
Last State: Terminated (Error)
Exit Code: 100
Events:
  Normal  Pulled  Successfully pulled image "trinodb/trino:475" in 127ms

exit code      100   (config validation)
OOMKilled      no
node pressure  none
quota events   none

trino-coordinator-68bf449659-thq8v   0/1  Running           2  2m7s
trino-worker-58f7b5d9dd-92mnn        0/1  CrashLoopBackOff  7  19m
trino-worker-58f7b5d9dd-bkgz6        1/1  Running           0  33m
trino-worker-58f7b5d9dd-qnv9t        0/1  CrashLoopBackOff  6  11m
trino-worker-58f7b5d9dd-r6tx6        0/1  CrashLoopBackOff  6  11m

2026-02-25T20:42:15.224Z ERROR main io.trino.server.Server Configuration errors:
1) Error: Connector 'bigquery' requires additional JVM argument(s).
   Please add the following to the JVM configuration:
   '--add-opens=java.base/java.nio=ALL-UNNAMED'

news_bq.properties:
  connector.name=bigquery
  project-id=********
  credentials-key=${ENV:bigquery_credentials_key}

jvm.config: |
  -Xmx8G
  -XX:+UseG1GC
  -XX:+ExitOnOutOfMemoryError
  # --add-opens=java.base/java.nio=ALL-UNNAMED  NOT present

trino-coordinator-68bf449659  CREATED 2026-02-25T20:08:47Z  REPLICAS 1
trino-worker-58f7b5d9dd       CREATED 2026-02-25T20:08:47Z  REPLICAS 4
worker lost minimum availability  20:35:54Z
alert fired                       20:43:31Z

2026-02-25T20:48:55.349Z ERROR main io.trino.server.Server Configuration errors:
1) Error: Connector 'bigquery' requires additional JVM argument(s).
   '--add-opens=java.base/java.nio=ALL-UNNAMED'

Final RCA

Remediation

# Option 1: Patch ConfigMaps directly
kubectl get configmap trino-worker -n analytics -o yaml | \
  sed 's/jvm.config: |/jvm.config: |\n  --add-opens=java.base\/java.nio=ALL-UNNAMED/' | \
  kubectl apply -f -

# Repeat for coordinator ConfigMap
kubectl get configmap trino-coordinator -n analytics -o yaml | \
  sed 's/jvm.config: |/jvm.config: |\n  --add-opens=java.base\/java.nio=ALL-UNNAMED/' | \
  kubectl apply -f -

# Option 2 (preferred): Update Helm values and redeploy
# In values.yaml under jvm.config, add:
#   --add-opens=java.base/java.nio=ALL-UNNAMED

# Restart pods to pick up config change
kubectl rollout restart deployment/trino-coordinator -n analytics
kubectl rollout restart deployment/trino-worker -n analytics

• Add --add-opens=java.base/java.nio=ALL-UNNAMED to Trino JVM configuration (Helm values preferred over direct ConfigMap edit).
• Add a CI/CD validation step that checks connector requirements against JVM configuration before deploying.
• Review Trino upgrade notes when bumping versions: connector JVM requirements change across versions.
• Consider adding a Helm chart test or init container that validates Trino configuration before the main container starts.
• Tune the Grafana alert to distinguish between ImagePull failures and application startup failures (Exit Code 100 != ImagePull).

The BigQuery connector's JVM-argument requirement was added to the catalog configuration without a matching update to jvm.config, so the gap sat invisible until the next rollout applied it cluster-wide. Moving jvm.config into the same Helm values file that enables a connector, so the two change together, removes the chance of shipping one without the other. A CI validation step that checks each enabled connector's documented JVM requirements against the rendered jvm.config, and a Helm chart test or init container that runs Trino's own config validation before the main container starts, would have caught this before it ever reached the cluster.

Frequently asked questions