Traces

Distributed tracing to analyze latency, detect slow spans, and find N+1 queries

Cased Telemetry captures distributed traces via the Sentry SDK or OpenTelemetry (through cased-agent). Analyze latency, find slow spans, detect N+1 query patterns, and track performance regressions.

Quick Start

# Find slow spans (>500ms)
cased perf slow --since 1h

# View latency percentiles
cased perf latency --since 1h --service api

# Detect N+1 query patterns
cased perf n1 --since 1h

# Check for regressions
cased perf regression --service api

Commands

Find Slow Spans

Identify spans that exceed a latency threshold:

# Find spans slower than 500ms (default)
cased perf slow --since 1h

# Custom threshold (1 second)
cased perf slow --since 24h --threshold 1000

# Filter by service
cased perf slow --service api --since 1h

Options:

Option	Description	Default
--since	Time range (1h, 24h, 7d)	1h
--threshold	Minimum duration in ms	500
--service	Filter by service name	-
--limit	Max results	50

Latency Percentiles

View p50, p95, p99 latencies grouped by service or endpoint:

# Overall latency by service
cased perf latency --since 1h

# Group by endpoint
cased perf latency --service api --group-by endpoint

# Both service and endpoint
cased perf latency --since 24h --group-by both

Output:

Service          p50      p95      p99      Count
─────────────────────────────────────────────────
api              45ms     120ms    350ms    12,450
worker           230ms    890ms    1.2s     3,200
gateway          12ms     35ms     78ms     45,000

Options:

Option	Description	Default
--since	Time range	1h
--service	Filter by service	-
--group-by	service, endpoint, or both	service

Detect N+1 Queries

Find repeated similar operations that indicate N+1 query patterns:

# Find N+1 patterns
cased perf n1 --since 1h

# Require more repetitions to flag
cased perf n1 --min-count 10

Output:

Pattern                              Count    Trace Example
────────────────────────────────────────────────────────────
SELECT * FROM users WHERE id = ?     47       abc123
SELECT * FROM orders WHERE user_id   23       def456

Options:

Option	Description	Default
--since	Time range	1h
--min-count	Minimum repetitions to flag	5

Trace Breakdown

Get a detailed service breakdown for a specific trace:

cased perf breakdown <trace_id>

Output:

Trace: abc123def456
Total Duration: 1.2s

Service          Duration    % of Total    Spans
──────────────────────────────────────────────────
database         680ms       56.7%         12
api              320ms       26.7%         3
redis            150ms       12.5%         8
external-api     50ms        4.2%          1

Detect Regressions

Compare recent performance against a baseline period:

# Compare last day to last week
cased perf regression --service api

# Custom periods
cased perf regression --service api --baseline 7d --compare 1d

# Filter by endpoint
cased perf regression --service api --endpoint /api/users

Output:

Endpoint              Baseline p95    Current p95    Change
────────────────────────────────────────────────────────────
/api/users            45ms            120ms          +167% ⚠️
/api/orders           230ms           245ms          +7%
/api/health           5ms             5ms            0%

Options:

Option	Description	Default
--service	Service to analyze (required)	-
--endpoint	Filter by endpoint	-
--baseline	Baseline period	7d
--compare	Comparison period	1d

Performance Summary

Get an overall summary of system performance:

cased perf summary --since 1h

Output:

Performance Summary (last 1h)
─────────────────────────────
Total Traces:     45,230
Total Spans:      234,500
Error Rate:       0.3%

Latency (all services):
  p50:  34ms
  p95:  180ms
  p99:  450ms

Slowest Services:
  1. database      avg: 89ms
  2. external-api  avg: 67ms
  3. worker        avg: 45ms

API Endpoints

All performance data is also available via REST API:

Slow Spans

curl -H "Authorization: Token YOUR_API_KEY" \
  "https://app.cased.com/api/v1/telemetry/traces/slow?since=1h&threshold=500"

Latency Percentiles

curl -H "Authorization: Token YOUR_API_KEY" \
  "https://app.cased.com/api/v1/telemetry/traces/latency?since=1h&group_by=service"

N+1 Detection

curl -H "Authorization: Token YOUR_API_KEY" \
  "https://app.cased.com/api/v1/telemetry/traces/n1?since=1h&min_count=5"

Investigation Workflow

A typical performance investigation workflow:

1. Start with summary to get an overview:

   cased perf summary --since 1h

2. Check for regressions if latency increased:

   cased perf regression --service api

3. Find slow spans to identify bottlenecks:

   cased perf slow --service api --since 1h

4. Check for N+1 queries if database is slow:

   cased perf n1 --since 1h

5. Drill into a specific trace for details:

   cased perf breakdown <trace_id>

Alerting

Set up alerts based on performance thresholds using Cased workflows.