Monitor Workloads

Monitor Workloads

Monitor Teradata workload definitions, activation status, query distribution, and TASM statistics using real-time MCP resources to understand how workload management is functioning and identify optimization opportunities.

🔍 Enhanced Capabilities

This skill now leverages real-time workload monitoring resources!

With tdwm-mcp v1.5.0, this skill provides:

• ✅ REAL-TIME WORKLOAD DATA - Instant workload distribution without queries
• ✅ CLASSIFICATION ANALYSIS - Understand how queries are routed to workloads
• ✅ TASM EFFECTIVENESS TRACKING - Monitor rule performance and statistics
• ✅ CONFIGURATION DISCOVERY - Explore filters, throttles, and rules per workload
• ✅ INTEGRATED METRICS - Workloads + queries + resources in one view

Instructions

When to Use This Skill

• User asks about workload configuration or status
• Need to understand how queries are being classified
• Investigating workload distribution or rule effectiveness
• Reviewing TASM performance and statistics
• Identifying misclassification issues
• Assessing workload balance and priority effectiveness

Available MCP Tools

Workload Monitoring:

• list_active_WD - List currently active workload definitions
• list_WDs - List all workloads (active and inactive)
• show_tdwm_summary - Display workload distribution dashboard
• show_tasm_statistics - View TASM performance metrics
• show_tasm_even_history - Review TASM event history and classification decisions

Related Analysis:

• list_query_band - See query bands used for classification
• list_delayed_request - View queries delayed by throttles
• show_trottle_statistics - Throttle impact per workload

Available MCP Resources (NEW ✨)

Real-Time Workload Data:

• tdwm://system/workloads - Active workload list with status
• tdwm://system/summary - Workload distribution and resource usage
• tdwm://system/delayed-queries - Queries delayed by workload
• tdwm://system/throttle-statistics - Throttle impact per workload

Configuration Discovery:

• tdwm://rulesets - List all available rulesets
• tdwm://system/active-ruleset - Currently active ruleset
• tdwm://ruleset/{name}/filters - Filters routing to workloads
• tdwm://ruleset/{name}/throttles - Throttles protecting workloads

Reference:

• tdwm://reference/workload-states - Workload state meanings
• tdwm://reference/classification-flow - How TASM classifies queries

Step-by-Step Workflow

Phase 1: Quick Assessment (Use Resources First)

1. Get Real-Time Workload Overview

   • Read resource: tdwm://system/workloads
   • Shows all active workloads with status
   • Provides quick inventory of workload configuration

2. Check Workload Distribution

   • Read resource: tdwm://system/summary
   • Shows how queries and resources are distributed
   • Identifies which workloads are busy
   • Provides utilization percentages

Phase 2: Detailed Analysis (Use Tools)

3. Review Workload Configuration

   • Use list_active_WD to see enabled workloads with details
   • Use list_WDs to see all workloads (active + inactive)
   • Understand workload hierarchy and priorities
   • Document workload purposes

4. Analyze Workload Distribution

   • Use show_tdwm_summary for detailed distribution data
   • See query counts, concurrency, resource usage per workload
   • Identify which workloads are handling most traffic
   • Compare against expected distribution

5. Check TASM Statistics

   • Use show_tasm_statistics to view performance metrics
   • Review rule activations, throttle actions, exceptions
   • Identify patterns in workload management actions
   • Calculate rule effectiveness rates

6. Review TASM Events

   • Use show_tasm_even_history for classification decisions
   • Analyze why queries landed in specific workloads
   • Look for unexpected or problematic patterns
   • Identify misclassification root causes

Phase 3: Configuration Exploration

7. Discover Workload Filters

   • Get active ruleset: tdwm://system/active-ruleset
   • List filters: tdwm://ruleset/{name}/filters
   • For each filter, check which workload it routes to
   • Identify gaps in coverage (queries not matching any filter)

8. Discover Workload Throttles

   • List throttles: tdwm://ruleset/{name}/throttles
   • Check which throttles protect which workloads
   • Review throttle limits and current usage
   • Identify workloads without protection

9. Assess Effectiveness

   • Determine if workloads are being used as intended
   • Check if classification rules are working correctly
   • Verify resource distribution matches priorities
   • Identify workloads that need tuning

Examples

Example 1: Quick Workload Status Check (Fast)

Scenario: "Show me the current workload configuration"

Action (Resource-First Approach):

1. Read active workloads:
   tdwm://system/workloads
   → PRODUCTION_HIGH (active, priority 1)
   → ETL_BATCH (active, priority 2)
   → INTERACTIVE (active, priority 2)
   → ANALYTICS (active, priority 3)
   → AD_HOC (active, priority 4)
   → DEFAULT (active, priority 5)
   → YEAR_END (inactive - seasonal)
   → MAINTENANCE (inactive - special purpose)

2. Check distribution:
   tdwm://system/summary
   → PRODUCTION_HIGH: 12 queries, 25% resources
   → ETL_BATCH: 45 queries, 45% resources
   → INTERACTIVE: 28 queries, 20% resources
   → ANALYTICS: 8 queries, 7% resources
   → AD_HOC: 5 queries, 3% resources
   → DEFAULT: 2 queries, <1% resources

3. Report:
   "Current Workload Configuration:

    Active Workloads (6):
    - PRODUCTION_HIGH: Priority 1, 12 queries, 25% resources
    - ETL_BATCH: Priority 2, 45 queries, 45% resources (dominant)
    - INTERACTIVE: Priority 2, 28 queries, 20% resources
    - ANALYTICS: Priority 3, 8 queries, 7% resources
    - AD_HOC: Priority 4, 5 queries, 3% resources
    - DEFAULT: Priority 5, 2 queries, <1% resources (minimal, good)

    Inactive Workloads (2):
    - YEAR_END: Seasonal (activate December-January)
    - MAINTENANCE: Special purpose (activate during maintenance windows)

    Assessment:
    ✅ ETL batch is dominant workload (expected for current time)
    ✅ DEFAULT has minimal queries (good classification)
    ✅ Resource distribution aligns with priorities"

Result: Complete workload inventory in seconds

Example 2: Investigate Workload Distribution Imbalance

Scenario: "Why is DEFAULT workload so busy? It should be nearly empty."

Action (Distribution Analysis):

1. Check current distribution:
   tdwm://system/summary
   → DEFAULT: 35 queries, 18% resources (⚠️ TOO HIGH!)
   → Expected: <5% resources

2. Get detailed summary:
   show_tdwm_summary()
   → DEFAULT workload details:
     - 35 active queries
     - Average runtime: 12 minutes
     - No throttle applied
     - Using 18% of system resources

3. Review TASM classification history:
   show_tasm_even_history(workload="DEFAULT")
   → Last 50 events show:
     - 65% "No matching filter" (no rule matched)
     - 35% "NEW_BI_TOOL query band not recognized"

4. Check query bands:
   list_query_band(type="TRANSACTION")
   → Filter for DEFAULT workload sessions
   → Common query bands:
     - APP=NEW_BI_TOOL (20 queries)
     - No query band set (15 queries)

5. Identify users:
   list_sessions()
   → Filter for DEFAULT workload
   → 20 queries from bi_users (NEW_BI_TOOL application)
   → 15 queries from adhoc_users (no query band)

6. Check existing filters:
   tdwm://ruleset/Tactical/filters
   → No filter matches APP=NEW_BI_TOOL
   → Ad-hoc users expected to not set query band

7. Report:
   "DEFAULT Workload Analysis: 35 queries (18% resources)

    Root Causes:
    1. NEW_BI_TOOL (20 queries, 57% of DEFAULT):
       → New BI application deployed last week
       → Sets query band 'APP=NEW_BI_TOOL'
       → No filter exists to route these queries
       → ACTION REQUIRED: Create filter for NEW_BI_TOOL

    2. Ad-hoc users (15 queries, 43% of DEFAULT):
       → No query band set (expected)
       → ACTION: Acceptable, or educate users

    Impact:
    - 18% of resources going to DEFAULT (should be <5%)
    - NEW_BI_TOOL queries not getting appropriate priority
    - Potential SLA violations for BI users

    RECOMMENDATION:
    Use tune-workloads skill to add classification:
    - Create or update filter to match APP=NEW_BI_TOOL
    - Route to INTERACTIVE workload or create dedicated BI workload
    - Will reduce DEFAULT from 18% to <5% resources"

Result: Root cause identified, specific action recommended

Example 3: TASM Performance Review (Effectiveness)

Scenario: "Is TASM working effectively? Are rules firing as expected?"

Action (TASM Analysis):

1. Get TASM statistics:
   show_tasm_statistics()
   → Total requests processed: 1,847 (last hour)
   → Filter actions: 1,802 (97.6% classification rate)
   → Throttle delays: 45 (2.4% delayed)
   → Exceptions: 3 (0.16% errors)

2. Break down by workload:
   show_tasm_statistics(group_by="workload")
   → PRODUCTION_HIGH: 245 requests, 0 delays
   → ETL_BATCH: 892 requests, 12 delays (throttled)
   → INTERACTIVE: 467 requests, 8 delays
   → ANALYTICS: 156 requests, 15 delays (throttled)
   → AD_HOC: 84 requests, 10 delays (throttled)
   → DEFAULT: 3 requests (excellent!)

3. Review recent TASM events:
   show_tasm_even_history(time_range="last_1_hour")
   → Filter events: 1,802 successful classifications
   → Throttle events: 45 delay queue placements
   → No classification: 3 (went to DEFAULT)

4. Check throttle effectiveness:
   tdwm://system/throttle-statistics
   → ETL_BATCH throttle: Limit 20, delayed 12 (working)
   → ANALYTICS throttle: Limit 10, delayed 15 (aggressive)
   → AD_HOC throttle: Limit 5, delayed 10 (aggressive)

5. Analyze delays:
   list_delayed_request()
   → 8 queries currently in delay queue
   → Average delay time: 2.3 minutes (acceptable)
   → Max delay time: 8.7 minutes (one query)

6. Report:
   "TASM Performance Review:

    Overall Effectiveness: ✅ EXCELLENT
    - 97.6% classification rate (target >95%)
    - Only 3 queries in DEFAULT (0.16%)
    - 2.4% queries throttled (working as designed)

    Workload Breakdown:
    - PRODUCTION_HIGH: 245 requests, 0 delays (protected priority)
    - ETL_BATCH: 892 requests, 12 delays (1.3% delay rate, healthy)
    - INTERACTIVE: 467 requests, 8 delays (1.7% delay rate, healthy)
    - ANALYTICS: 156 requests, 15 delays (9.6% delay rate, aggressive)
    - AD_HOC: 84 requests, 10 delays (11.9% delay rate, aggressive)

    Throttle Effectiveness:
    ✅ ETL throttle working well (1.3% delay rate)
    ✅ Interactive protected (1.7% delay rate)
    ⚠️  Analytics throttle may be too restrictive (9.6% delay rate)
    ⚠️  Ad-hoc throttle may be too restrictive (11.9% delay rate)

    Recommendations:
    1. Monitor ANALYTICS and AD_HOC workloads
       → 9-12% delay rates higher than ideal (<5%)
       → Consider increasing throttle limits slightly

    2. Investigate 3 DEFAULT queries
       → Find why they didn't classify
       → Add filters if needed

    3. TASM is performing well overall
       → No major configuration changes needed"

Result: Comprehensive effectiveness assessment with specific recommendations

Example 4: Discover Workload Configuration (Exploration)

Scenario: "What filters and throttles are configured for the ETL_BATCH workload?"

Action (Configuration Discovery):

1. Get active ruleset:
   tdwm://system/active-ruleset
   → "Tactical"

2. List all filters:
   tdwm://ruleset/Tactical/filters
   → Found 8 filters

3. Find ETL_BATCH filter:
   tdwm://ruleset/Tactical/filter/ETL_FILTER
   → Name: ETL_FILTER
   → Action: ACCEPT
   → Workload: ETL_BATCH
   → Enabled: true
   → Classifications:
     - APPL = 'ETL_*' (Inclusion)
     - USER = 'etl_batch' (Inclusion)

4. List all throttles:
   tdwm://ruleset/Tactical/throttles
   → Found 6 throttles

5. Find ETL_BATCH throttle:
   tdwm://ruleset/Tactical/throttle/ETL_LIMIT
   → Name: ETL_LIMIT
   → Type: DM (Delay Management)
   → Limit: 20 concurrent queries
   → Enabled: true
   → Classifications:
     - APPL = 'ETL_*' (Inclusion)

6. Check current utilization:
   tdwm://system/summary
   → ETL_BATCH workload:
     - 18 active queries (under 20 limit)
     - No queries delayed currently
     - 45% of system resources

7. Review classification effectiveness:
   show_tasm_even_history(workload="ETL_BATCH")
   → All ETL queries correctly classified
   → Filter matching on APPL and USER working

8. Report:
   "ETL_BATCH Workload Configuration:

    Filter (ETL_FILTER):
    - Routes queries to ETL_BATCH workload
    - Matches queries where:
      → Application starts with 'ETL_' (query band), OR
      → Username is 'etl_batch'
    - Status: Enabled and working correctly

    Throttle (ETL_LIMIT):
    - Limits ETL_BATCH to 20 concurrent queries
    - Type: Delay Management (queues excess queries)
    - Status: Enabled, currently under limit (18/20)

    Current State:
    - 18 active queries (2 slots available)
    - 0 queries delayed (throttle not engaged)
    - 45% of system resources (dominant workload)

    Classification Effectiveness:
    ✅ All ETL queries correctly classified
    ✅ No misclassifications detected
    ✅ Filter working as designed

    Configuration Assessment:
    ✅ Properly configured with both filter and throttle
    ✅ Throttle limit appropriate for current load
    ✅ Classification criteria catching all ETL queries"

Result: Complete configuration documentation for workload

Example 5: Compare Workload Design vs Actual Usage (Audit)

Scenario: "Are our workloads being used as originally designed?"

Action (Design Validation):

1. Document design intent (from requirements):
   - PRODUCTION_HIGH: Critical production apps, <10% queries, 30% resources
   - ETL_BATCH: Nightly ETL, 40% queries, 40% resources
   - INTERACTIVE: User dashboards, 30% queries, 20% resources
   - ANALYTICS: Analytics queries, 15% queries, 8% resources
   - AD_HOC: Ad-hoc exploration, 5% queries, 2% resources
   - DEFAULT: Miscellaneous, <1% queries, <1% resources

2. Get actual distribution:
   tdwm://system/summary
   ACTUAL:
   - PRODUCTION_HIGH: 12% queries, 25% resources ✅ Close
   - ETL_BATCH: 45% queries, 45% resources ✅ Close
   - INTERACTIVE: 28% queries, 20% resources ✅ Match!
   - ANALYTICS: 8% queries, 7% resources ✅ Match!
   - AD_HOC: 5% queries, 3% resources ⚠️  Slightly high
   - DEFAULT: 2% queries, <1% resources ⚠️  Should be <1%

3. Analyze deviations:

   PRODUCTION_HIGH (12% vs designed 10%):
   - show_tasm_even_history(workload="PRODUCTION_HIGH")
   → Slightly more production apps than expected
   → Within acceptable range

   AD_HOC (5% queries, 3% resources vs designed 5% queries, 2% resources):
   - show_query_log(workload="AD_HOC")
   → Ad-hoc queries running longer than expected
   → Using more resources per query

   DEFAULT (2% vs designed <1%):
   - show_tasm_even_history(workload="DEFAULT")
   → NEW_BI_TOOL queries not classified (need filter)

4. Check SLA compliance:
   - PRODUCTION_HIGH: 99.9% uptime requirement
     → show_query_log(workload="PRODUCTION_HIGH", calc="avg_response_time")
     → Average: 0.8 seconds (target <1 second) ✅

   - INTERACTIVE: 95th percentile <5 seconds
     → show_query_log(workload="INTERACTIVE", calc="p95_response_time")
     → 95th percentile: 4.2 seconds ✅

   - ETL_BATCH: Complete within 4-hour window
     → show_query_log(workload="ETL_BATCH", time_range="last_night")
     → Completed in 3.5 hours ✅

5. Report:
   "Workload Design vs Actual Usage Audit:

    Overall Assessment: 🟢 MOSTLY ALIGNED

    Workload Comparison:
    ✅ PRODUCTION_HIGH: 12% queries vs 10% design (acceptable)
    ✅ ETL_BATCH: 45% queries vs 40% design (acceptable)
    ✅ INTERACTIVE: 28% queries vs 30% design (perfect)
    ✅ ANALYTICS: 8% queries vs 15% design (underutilized)
    ⚠️  AD_HOC: 3% resources vs 2% design (slightly over)
    ⚠️  DEFAULT: 2% queries vs <1% design (needs attention)

    SLA Compliance:
    ✅ PRODUCTION_HIGH: 0.8s avg (<1s target)
    ✅ INTERACTIVE: 4.2s p95 (<5s target)
    ✅ ETL_BATCH: 3.5hrs window (<4hrs target)

    Issues Found:
    1. DEFAULT workload higher than design (2% vs <1%):
       → Root cause: NEW_BI_TOOL not classified
       → ACTION: Create filter for NEW_BI_TOOL

    2. AD_HOC using more resources than design (3% vs 2%):
       → Ad-hoc queries running longer than expected
       → ACTION: Review slow ad-hoc queries, consider optimization

    3. ANALYTICS underutilized (8% vs 15% design):
       → Less analytics usage than projected
       → ACTION: No immediate action, monitor trend

    Recommendations:
    - Fix DEFAULT classification issue (Priority 1)
    - Investigate AD_HOC resource usage (Priority 2)
    - Update design docs to reflect ANALYTICS lower usage (Priority 3)
    - Overall workload management is effective, minor tuning needed"

Result: Comprehensive design validation with prioritized actions

Best Practices

Resource-First Approach (NEW ✨)

• START with resources for instant workload overview (tdwm://system/workloads)
• Resources provide real-time snapshot without adding system load
• Use tools for detailed analysis and historical data
• Combine resources + tools for complete picture

Workload Inventory Management

• Active workloads define current system behavior - verify they match intent
• Inactive workloads should be documented (seasonal, emergency, retired)
• DEFAULT workload should have minimal traffic (<5% queries, <1% resources)
• Workload names should clearly indicate purpose and priority

Distribution Analysis

• Workload distribution should align with business priorities
• Compare actual distribution against design/SLA targets
• High DEFAULT workload usage indicates classification problems
• Resource % should roughly match priority rankings

TASM Effectiveness Monitoring

• Monitor TASM statistics regularly to catch rule issues early
• Target: >95% classification rate (queries not in DEFAULT)
• High exception counts indicate classification problems
• Track TASM statistics over time to identify trends

Configuration Discovery (NEW ✨)

• Use MCP resources to explore filters and throttles per workload
• Verify each workload has both filter (routing) and throttle (protection)
• Check classification criteria match expected query patterns
• Validate throttle limits against actual concurrency needs

Classification Troubleshooting

• DEFAULT workload is the catch-all - investigate why queries land there
• Use TASM event history to understand classification decisions
• Verify query bands are being set by applications
• Check filter criteria match actual query band values

SLA Validation

• Document SLA targets for each workload
• Regularly compare actual performance against SLA targets
• Use query logs to calculate response time percentiles
• Alert on SLA violations

Related Skills

• Use tune-workloads skill to fix classification issues
• Use manage-workloads skill to create filters/throttles for workloads
• Use optimize-throttles skill to adjust throttle limits
• Use discover-configuration skill for systematic configuration audit
• Use monitor-queries skill to correlate query patterns with workloads