fabric-spark-compute-remediate
Skill17 starsChanged 9 months agoMITView on GitHub
Diagnose and resolve Microsoft Fabric Spark compute issues including HTTP 430 throttling errors, job queueing failures, slow session startup, autoscale problems, capacity SKU limits, environment publishing failures, library conflicts, node sizing, burst factor configuration, and VNet provisioning delays. Use when remediate Spark notebooks, Spark job definitions, lakehouse operations, starter pools, custom pools, or concurrency limits in Fabric workspaces.
In this repository24 other files it gives its agents.
- fabric-data-agent-perf-remediateSkill
- fabric-data-agent-remediateSkill
- fabric-data-agentSkill
- fabric-data-factory-perf-remediateSkill
- fabric-dataflows-perf-remediationSkill
- fabric-delta-spark-perfSkill
- fabric-iqSkill
- fabric-lakehouse-access-controlSkill
- fabric-lakehouse-perf-remediateSkill
- fabric-lakehouse-views-perf-remediateSkill
- fabric-network-remediateSkill
- fabric-notebook-perf-remediateSkill
- fabric-onelake-perf-remediateSkill
- fabric-pandas-perf-remediateSkill
- fabric-pbi-perf-remediateSkill
- fabric-pbi-security-remediateSkill
- fabric-performance-monitoringSkill
- fabric-pyspark-perf-remediateSkill
- fabric-rest-api-perf-remediateSkill
- fabric-rest-api-remediateSkill
- fabric-rti-perf-remediateSkill
- fabric-spark-compute-perf-remediateSkill
- fabric-spark-perf-remediateSkill
- fabric-udf-perf-remediateSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 12 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈12
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 50
Every Fabric capacity SKU provides Spark VCores at a fixed ratio with a 3x burst factor:
- Distinctive: line 80
If Starter Pool with no custom libraries: expect 5-10 seconds.
- Distinctive: line 81
If custom libraries or Spark properties are attached via environment: expect 30 seconds to 5 minutes
- Distinctive: line 82
If using non-Medium node size: Starter Pool fast-start is unavailable.
- Distinctive: line 83
If Private Link is enabled and this is the first job: expect 10-15 minute VNet provisioning delay
- Distinctive: line 89
Check if another Publish action is already in progress (only one at a time)
- Distinctive: line 90
Verify library compatibility with the selected Spark runtime version
- Distinctive: line 91
If runtime was recently changed, remove incompatible libraries and republish
- Distinctive: line 92
If Private Link is enabled, the first publish may trigger VNet provisioning (10-15 min delay)
- Distinctive: line 99
Run the Spark capacity calculator to quickly determine VCore limits, max nodes, and queue limits for any Fabric SKU.
- Distinctive: line 113
Use Starter Pools when: you need fast startup (<10s), workloads fit Medium nodes (8 VCores, 64 GB), and you have no heavy library …
- Very specific: line 115
Use Custom Pools when: you need Large/X-Large/XX-Large nodes for memory-intensive workloads, you need precise control over min/max node …
- Distinctive: line 121
Disable: when: you have a multi-tenant environment with many concurrent users and fairness across teams matters more than single-job …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Microsoft Fabric Spark Compute remediate
Structured diagnostic workflows for resolving Apache Spark compute issues in Microsoft Fabric Data Engineering and Data Science workloads.
When to Use This Skill
- Spark jobs fail with HTTP 430 throttling or TooManyRequestsForCapacity errors
- Notebook or Spark job sessions are slow to start (>30 seconds)
- Environment publishing fails or hangs
- Autoscale is not scaling up or down as expected
- Jobs are queued indefinitely or expiring after 24 hours
- Custom pool creation fails or pools are undersized
- Library installation causes session startup delays
- Capacity appears exhausted despite low job counts
- VNet/Private Link provisioning adds unexpected delays
- Burst factor or job-level bursting behavior is unclear
Prerequisites
- Workspace Admin or Capacity Admin role in Microsoft Fabric
- Access to the Monitoring Hub for active Spark sessions
- Access to Workspace Settings > Data Engineering/Science
- Knowledge of current Fabric capacity SKU (F2 through F2048)
Quick Diagnostic: Identify Your Issue
Start here. Match your symptom to a diagnostic path:
| Symptom | Diagnostic Path |
| HTTP 430 error | See Throttling and Concurrency |
| Jobs stuck in queue | See Throttling and Concurrency |
| Slow session startup | See Session and Environment |
| Environment publish fails | See Session and Environment |
| Autoscale not working | See Pool Configuration |
| Pool sizing questions | See Pool Configuration |
| Library conflicts | See Session and Environment |
| VNet delay on first job | See Session and Environment |
Core Concepts
Capacity Unit to VCore Mapping
Every Fabric capacity SKU provides Spark VCores at a fixed ratio with a 3x burst factor:
1 Capacity Unit = 2 Spark VCores
For an F64 SKU: 64 CU x 2 = 128 base VCores, with 3x burst = 384 max Spark VCores.
Job Admission Model
Fabric Spark uses optimistic job admission: jobs are admitted based on their minimum core requirement (determined by the pool's minimum node setting). Jobs start with minimum nodes and scale up toward maximum nodes as cores become available. If no cores are available for the minimum requirement, the job is rejected or queued.
Two Pool Types
- Starter Pools: Pre-warmed, medium nodes only, 5-10 second startup, always available
- Custom Pools: User-configured node sizes (Small through XX-Large), 2-5 minute cold start, full flexibility
remediate Workflows
Workflow 1: HTTP 430 Throttling
- Confirm the error:
HTTP Response code 430: This Spark job can't be run because you have hit a Spark compute or API rate limit - Open the Monitoring Hub and count active Spark sessions
- Calculate your capacity's max VCores:
SKU CU × 2 × 3 (burst) = max VCores - Compare active usage against max VCores
- Resolve by canceling idle sessions, upgrading SKU, or enabling job queueing for pipeline/scheduler jobs
See throttling-and-concurrency.md for the full SKU limits table and queue configuration.
Workflow 2: Slow Session Startup
- Determine pool type (Starter vs Custom)
- If Starter Pool with no custom libraries: expect 5-10 seconds; if slower, check capacity utilization
- If custom libraries or Spark properties are attached via environment: expect 30 seconds to 5 minutes
- If using non-Medium node size: Starter Pool fast-start is unavailable; expect 2-5 minutes (on-demand)
- If Private Link is enabled and this is the first job: expect 10-15 minute VNet provisioning delay
See session-and-environment.md for detailed diagnosis.
Workflow 3: Environment Publishing Failure
- Check if another Publish action is already in progress (only one at a time)
- Verify library compatibility with the selected Spark runtime version
- If runtime was recently changed, remove incompatible libraries and republish
- If Private Link is enabled, the first publish may trigger VNet provisioning (10-15 min delay)
- Review the error notification for specific failure details
See session-and-environment.md for resolution steps.
Available Scripts
Run the Spark capacity calculator to quickly determine VCore limits, max nodes, and queue limits for any Fabric SKU.
# Calculate capacity for F64 SKU
./scripts/Get-FabricSparkCapacity.ps1 -SkuSize 64
# Compare multiple SKUs
./scripts/Get-FabricSparkCapacity.ps1 -SkuSize 64 -CompareWith 128,256Key Decision Points
When to Use Starter Pools vs Custom Pools
Use Starter Pools when: you need fast startup (<10s), workloads fit Medium nodes (8 VCores, 64 GB), and you have no heavy library dependencies.
Use Custom Pools when: you need Large/X-Large/XX-Large nodes for memory-intensive workloads, you need precise control over min/max node counts, or you need to limit autoscale behavior.
When to Enable Job-Level Bursting
Enable (default) when: you run large batch jobs that benefit from consuming all available burst VCores and concurrency is low.
Disable when: you have a multi-tenant environment with many concurrent users and fairness across teams matters more than single-job throughput.
Admin Portal path: Capacity Settings > Data Engineering/Science > Disable Job-Level Bursting toggle.
References
- Throttling and Concurrency Guide — SKU limits, queue sizes, HTTP 430 resolution
- Session and Environment Guide — Startup delays, publishing, libraries, VNet
- Pool Configuration Guide — Node sizing, autoscale, custom pool setup, billing
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

