Databricks Common Errors

Overview

Quick-reference diagnostic guide for the most frequent Databricks errors. Covers cluster failures, Spark OOM, Delta Lake conflicts, permissions, schema mismatches, rate limits, and job run failures with real SDK/SQL solutions.

Prerequisites

• Databricks CLI configured
• Access to cluster/job logs
• databricks-sdk installed for programmatic debugging

Instructions

Step 1: Identify the Error Source

# Get failed run details
databricks runs get --run-id $RUN_ID --output json | jq '{
  state: .state.result_state,
  message: .state.state_message,
  tasks: [.tasks[] | {key: .task_key, state: .state.result_state, error: .state.state_message}]
}'

Step 2: Match and Fix

CLUSTERNOTREADY / INVALID_STATE

ClusterNotReadyException: Cluster 0123-456789-abcde is not in a RUNNING state

Cause: Cluster is starting, terminating, or in error state.

from databricks.sdk import WorkspaceClient
from databricks.sdk.service.compute import State

w = WorkspaceClient()
cluster = w.clusters.get(cluster_id="0123-456789-abcde")

if cluster.state in (State.PENDING, State.RESTARTING):
    w.clusters.ensure_cluster_is_running("0123-456789-abcde")
elif cluster.state == State.TERMINATED:
    w.clusters.start_and_wait(cluster_id="0123-456789-abcde")
elif cluster.state == State.ERROR:
    reason = cluster.termination_reason
    print(f"Cluster error: {reason.code} — {reason.parameters}")
    # Common: CLOUD_PROVIDER_LAUNCH_FAILURE, INSTANCE_POOL_CLUSTER_FAILURE

SPARKDRIVEROOM

java.lang.OutOfMemoryError: Java heap space
SparkException: Job aborted due to stage failure

Cause: Driver or executor running out of memory.

# Fix 1: Increase memory via cluster Spark config
spark_conf = {
    "spark.driver.memory": "8g",
    "spark.executor.memory": "8g",
    "spark.sql.shuffle.partitions": "400",  # reduce skew
}

# Fix 2: Never collect() large datasets
# BAD:  all_data = df.collect()
# GOOD: df.write.format("delta").saveAsTable("catalog.schema.results")

# Fix 3: Broadcast small tables instead of shuffling
from pyspark.sql.functions import broadcast
result = large_df.join(broadcast(small_lookup_df), "key")

DELTACONCURRENTWRITE

ConcurrentAppendException: Files were added by a concurrent update
ConcurrentDeleteReadException: A concurrent operation modified files

Cause: Multiple jobs writing to the same Delta table simultaneously.

from delta.tables import DeltaTable
import time

def merge_with_retry(spark, source_df, target_table, merge_key, max_retries=3):
    """MERGE with retry for concurrent write conflicts."""
    for attempt in range(max_retries):
        try:
            target = DeltaTable.forName(spark, target_table)
            (target.alias("t")
                .merge(source_df.alias("s"), f"t.{merge_key} = s.{merge_key}")
                .whenMatchedUpdateAll()
                .whenNotMatchedInsertAll()
                .execute())
            return
        except Exception as e:
            if "Concurrent" in str(e) and attempt < max_retries - 1:
                time.sleep(2 ** attempt)
                continue
            raise

PERMISSION_DENIED

PERMISSION_DENIED: User does not have SELECT on TABLE catalog.schema.table
PermissionDeniedException: User does not have permission MANAGE on cluster

Cause: Missing Unity Catalog grants or workspace permissions.

-- Fix Unity Catalog permissions (requires GRANT privilege)
GRANT USAGE ON CATALOG analytics TO `data-team`;
GRANT USAGE ON SCHEMA analytics.silver TO `data-team`;
GRANT SELECT ON TABLE analytics.silver.orders TO `data-team`;

-- Check current grants
SHOW GRANTS ON TABLE analytics.silver.orders;

# Fix workspace object permissions
databricks permissions update jobs --job-id 123 --json '{
  "access_control_list": [{
    "user_name": "user@company.com",
    "permission_level": "CAN_MANAGE_RUN"
  }]
}'

INVALIDPARAMETERVALUE

InvalidParameterValue: Instance type xyz not supported in region us-east-1
Invalid spark_version: 13.x.x-scala2.12

Cause: Wrong cluster config for the workspace region.

w = WorkspaceClient()

# List valid node types for this workspace
for nt in sorted(w.clusters.list_node_types().node_types, key=lambda x: x.memory_mb)[:10]:
    print(f"{nt.node_type_id}: {nt.memory_mb}MB, {nt.num_cores} cores")

# List valid Spark versions
for v in w.clusters.spark_versions().versions:
    if "LTS" in v.name:
        print(f"{v.key}: {v.name}")

SCHEMA_MISMATCH

AnalysisException: A schema mismatch detected when writing to the Delta table

Cause: Source schema doesn't match target table.

# Option 1: Enable schema evolution
df.write.format("delta").option("mergeSchema", "true").mode("append").saveAsTable("target")

# Option 2: Identify differences
source_cols = set(df.columns)
target_cols = set(spark.table("target").columns)
print(f"Missing in source: {target_cols - source_cols}")
print(f"Extra in source: {source_cols - target_cols}")

# Option 3: Cast to match target schema
target_schema = spark.table("target").schema
for field in target_schema:
    if field.name in df.columns:
        df = df.withColumn(field.name, col(field.name).cast(field.dataType))

JOBRUNFAILED

RunState: FAILED — Run terminated with error

w = WorkspaceClient()
run = w.jobs.get_run(run_id=12345)

print(f"State: {run.state.life_cycle_state}")
print(f"Result: {run.state.result_state}")
print(f"Message: {run.state.state_message}")

# Check each task
for task in run.tasks:
    if task.state.result_state and task.state.result_state.value == "FAILED":
        output = w.jobs.get_run_output(task.run_id)
        print(f"Task '{task.task_key}' failed: {output.error}")
        if output.error_trace:
            print(f"Traceback:\n{output.error_trace[:500]}")

HTTP 429 — RATELIMITEXCEEDED

See databricks-rate-limits skill for full retry patterns.

from databricks.sdk.errors import TooManyRequests
import time

def call_with_backoff(operation, max_retries=5):
    for attempt in range(max_retries):
        try:
            return operation()
        except TooManyRequests as e:
            wait = e.retry_after_secs or (2 ** attempt)
            print(f"Rate limited, waiting {wait}s...")
            time.sleep(wait)
    raise RuntimeError("Max retries exceeded")

Output

• Error identified and categorized
• Fix applied from matching error pattern
• Resolution verified

Error Handling

Examples

Quick Diagnostic Commands

databricks clusters get --cluster-id $CID | jq '{state, termination_reason}'
databricks runs list --job-id $JID --limit 5 | jq '.runs[] | {run_id, state: .state.result_state}'
databricks permissions get jobs --job-id $JID

Escalation Path

1. Check Databricks Status
2. Collect evidence with databricks-debug-bundle
3. Search Community Forum
4. Contact support with workspace ID and request ID from error response

Resources

• Troubleshooting Guide
• Delta Lake Troubleshooting
• Resource Limits

Next Steps

For comprehensive debugging, see databricks-debug-bundle.