Module: NatsWork::WorkerHealth
- Defined in:
- lib/natswork/health_check.rb
Overview
Worker-specific health checks
Class Method Summary collapse
Class Method Details
.setup_worker_checks(worker, health_checker = HealthChecker.global) ⇒ Object
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 |
# File 'lib/natswork/health_check.rb', line 200 def self.setup_worker_checks(worker, health_checker = HealthChecker.global) # Worker status check health_checker.add_check('worker_status') do if worker.running? if worker.paused? { status: :degraded, message: 'Worker is paused' } elsif worker.stopping? { status: :degraded, message: 'Worker is stopping' } else { status: :healthy, message: 'Worker running normally' } end else { status: :unhealthy, message: 'Worker is stopped' } end end # Active jobs check health_checker.add_check('active_jobs') do stats = worker.stats active = stats[:active_jobs] concurrency = stats[:concurrency] if active >= concurrency { status: :degraded, message: "All #{concurrency} job slots occupied" } elsif active > concurrency * 0.8 { status: :degraded, message: "#{active}/#{concurrency} job slots occupied" } else { status: :healthy, message: "#{active}/#{concurrency} job slots occupied" } end end # Job processing rate check health_checker.add_check('job_processing') do stats = worker.stats processed = stats[:jobs_processed] failed = stats[:jobs_failed] if processed.zero? { status: :unknown, message: 'No jobs processed yet' } else error_rate = failed.to_f / processed if error_rate > 0.5 { status: :unhealthy, message: "High error rate: #{(error_rate * 100).round(1)}%" } elsif error_rate > 0.1 { status: :degraded, message: "Elevated error rate: #{(error_rate * 100).round(1)}%" } else { status: :healthy, message: "Error rate: #{(error_rate * 100).round(1)}%" } end end end end |