Partitions View

The Partitions view provides comprehensive partition monitoring with queue analytics, efficiency metrics, and wait time analysis.

Partitions view showing queue depth, efficiency metrics, and analytics

Overview

The Partitions view displays all SLURM partitions with real-time queue information, resource allocation, and performance metrics. Use this view to understand partition utilization and identify bottlenecks.

Table Columns

The partitions table displays 11 columns:

Queue Depth Visualization

The Queue Depth column shows a visual bar representing job distribution:

Copy
Queue: ████████░░░░
       ^^^^^^^^       Running jobs (solid)
               ░░░░   Pending jobs (hollow)

Visual indicators:

• █ (solid) = Running jobs
• ░ (hollow) = Pending jobs
• Length = Relative proportion

Color coding:

• Green = Healthy (low queue depth)
• Yellow = Moderate queue
• Red = High queue depth (bottleneck)

Partition States

Efficiency Metrics

The Efficiency column shows partition resource utilization:

85% [████████████████░░]

Rating levels:

• 90-100% (Green) - Excellent utilization
• 75-89% (Yellow) - Good utilization
• 60-74% (Orange) - Fair utilization
• <60% (Red) - Poor utilization

Efficiency calculation:

Efficiency = (Allocated Resources / Total Resources) × Job Success Rate

Factors considered:

• CPU allocation ratio
• Memory allocation ratio
• Job completion rate
• Resource waste from failed jobs

Partition Actions

View Partition Details

Shortcut: Enter

Shows comprehensive partition information:

Basic Information:

• Partition name and state
• Default/maximum time limits
• Node count and CPU count
• Priority level

Resource Limits:

• Maximum nodes per job
• Maximum CPUs per job
• Maximum memory per job
• Default resources

Access Control:

• Allowed accounts
• Allowed QoS levels
• User access restrictions

Current Status:

• Running jobs count
• Pending jobs count
• Allocated vs. total resources
• Current efficiency rating

View Partition Jobs

Shortcut: J

Switches to Jobs view filtered by this partition. Shows all jobs (running and pending) assigned to the partition.

View Partition Nodes

Shortcut: N

Switches to Nodes view filtered by this partition. Shows all nodes belonging to the partition with their current states.

Partition Analytics

Shortcut: A

Opens detailed analytics dashboard for the partition:

Queue Analytics:

• Job submission rate trends
• Queue depth over time
• Wait time distribution
• Peak usage hours

Resource Analytics:

• CPU utilization trends
• Memory utilization trends
• Node availability over time
• Resource fragmentation analysis

Performance Metrics:

• Job throughput (jobs/hour)
• Average job duration
• Success rate percentage
• Failed job analysis

Capacity Planning:

• Projected growth
• Resource recommendations
• Bottleneck identification
• Optimization suggestions

Wait Time Analytics

Shortcut: W

Shows cluster-wide wait time analysis:

By Partition:

• Average wait time per partition
• Maximum wait times
• Wait time trends
• Comparison across partitions

By Job Size:

• Wait times for small jobs (<10 nodes)
• Wait times for medium jobs (10-100 nodes)
• Wait times for large jobs (>100 nodes)

Status Indicators:

• 🟢 Good (<1 hour average wait)
• 🟡 Moderate (1-4 hours average wait)
• 🔴 Poor (>4 hours average wait)

Historical Trends:

• Wait time changes over 24 hours
• Week-over-week comparison
• Seasonal patterns

Filtering & Search

Simple Text Filter

Shortcut: /

Filters partitions by:

• Partition name
• State
• QoS policies
• Any displayed column

Global Search

Shortcut: Ctrl+F

Search across all cluster resources.

Sorting

Sort partitions by clicking column headers or using number keys.

Common sorting:

• By efficiency (identify underutilized partitions)
• By queue depth (find bottlenecks)
• By wait time (prioritize improvements)
• By node count (understand capacity)

Press S to open the interactive sort modal.

Keyboard Shortcuts Reference

Partition Operations

Filtering & Search

Data Management

Partition Details Example

When viewing partition details (Enter):

Copy
Partition: gpu
State: UP
Priority: 100

Time Limits:
  Default: 2-00:00:00 (2 days)
  Maximum: 7-00:00:00 (7 days)

Resources:
  Nodes: 32
  CPUs: 1536 (48 per node)
  Memory: 8TB total

Limits:
  Max Nodes per Job: 16
  Max CPUs per Job: 768
  Max Memory per Job: 4TB

Access:
  Allowed Accounts: research, ml-team
  Allowed QoS: normal, high
  Default QoS: normal

Current Status:
  Running Jobs: 28
  Pending Jobs: 15
  Allocated CPUs: 1200/1536 (78%)
  Efficiency: 85% (Good)

Understanding Queue Metrics

Average Wait Time

The average time pending jobs wait before starting execution.

Factors affecting wait time:

• Partition capacity
• Job priority and QoS
• Resource request size
• Current queue depth
• Fair-share scheduling

Healthy ranges:

• <1 hour: Excellent responsiveness
• 1-4 hours: Good for production
• 4-12 hours: Acceptable for batch processing

• 12 hours: Investigate bottleneck

Maximum Wait Time

The longest current wait for any pending job.

When to investigate:

• Max wait >24 hours: Check for stuck jobs
• Max wait >> Avg wait: Potential scheduling issue
• Max wait increasing: Growing bottleneck

Efficiency Analysis

High Efficiency (90-100%)

Indicators:

• Green efficiency bar
• High resource utilization
• Low waste

Potential issues:

• May be over-subscribed
• Users might experience long waits
• Consider expanding capacity

Low Efficiency (<60%)

Indicators:

• Red efficiency bar
• Low resource utilization
• High waste

Common causes:

• Poor resource request sizing
• Job failures
• Partition misconfiguration
• Seasonal low usage

Actions:

• Review job submission patterns
• Educate users on right-sizing
• Consider consolidating partitions
• Check for hardware issues

Tips

• Monitor wait times: Use W to identify partitions needing attention
• Check efficiency regularly: Low efficiency may indicate training opportunities
• Use analytics: Press A for in-depth partition analysis
• Navigate to jobs: Press J to see exactly what's running on a partition
• Navigate to nodes: Press N to check partition node health
• Compare partitions: Sort by efficiency to compare partition utilization
• Queue depth visualization: Use visual bars to quickly spot bottlenecks
• Filter by state: Use / with UP to focus on active partitions
• Capacity planning: Review analytics regularly for growth trends

Capacity Planning

Use partition analytics to plan capacity:

1. Identify bottlenecks: High queue depth + long wait times
2. Find underutilized: Low efficiency + low queue depth
3. Analyze trends: Job submission patterns over time
4. Right-size partitions: Match partition size to demand
5. Balance resources: Distribute workload across partitions

Analytics dashboard (A) provides:

• Historical utilization trends
• Projected resource needs
• Optimization recommendations
• Cost-benefit analysis for expansion