Runningop-0144
Scale · image-resizer
production · S3 event burst
4 → 8 replicas
3s elapsed15s limit
Retain 14 days
Older events purge nightly
Timed
ScalerMOCK
AM
production / us-east-1
Scaling overview
Safe, focused control for native workloads and AWS SQS.
Simple by design
Native Kubernetes first · AWS SQS is the only external source · existing HPA ownership is protected.
Active replicas
38
6 pending
Control sources
2
Native Kubernetes + SQS
Median API accept
420ms
Pod readiness tracked separately
Ownership guards
2
HPA workloads are read-only
Live operations
Current and recent operations
Timed attempts stay immutable; re-runs and reverts create new audit records.
2 running1 timed out
Runningop-0142
Scale · events-worker
production · SQS policy
18 → 24 replicas
7s elapsed30s limit
Retain 30 days
Older events purge nightly
Timed
Succeeded Reversibleop-0141
Scale · pdf-renderer
jobs · AM
0 → 8 replicas
3s elapsed10s limit
Retain 90 days
Older events purge nightly
Timed out Critical Slack sentop-0140
Scale · events-worker
production · Schedule
SCALE_UP_TIMED_OUT
18 → 36 replicas
10s elapsed10s limit
Retain 14 days
Older events purge nightly
Scheduledop-0139
Schedule · pdf-renderer
jobs · Schedule · 23:30
8 → 0 replicas
Starts 23:3010s limit
Retain 30 days
Older events purge nightly
Workloads
Ownership is checked before every write
| Workload | Replicas | Signal | Owner | Last operation | State | Actions |
|---|---|---|---|---|---|---|
checkout-api production | 12/ 12 | CPU · 61% | Native HPA | 1.8s 18 sec ago | Stable | |
events-worker production | 18/ 24 | SQS · 8.2k messages | Scaler | 2.4s 18 sec ago | Scaling | |
recommendations production | 8/ 8 | RPS · 1.4k | Native HPA | 1.2s 18 sec ago | Stable | |
image-resizer production | 4/ 8 | S3 · uploads bucket | Scaler | 3.1s 18 sec ago | Scaling | |
fraud-alerts production | 3/ 3 | CW Alarm · backlog-high | Scaler | 1.6s 18 sec ago | Stable | |
ingest-pipeline jobs | 6/ 6 | CloudWatch · DynamoDB WCU | Scaler | 2.0s 18 sec ago | Stable | |
pdf-renderer jobs | 0/ 0 | Schedule · 23:30 | Scaler | — 18 sec ago | Paused |
Advisory
Add an SQS failure floor
Queue metrics failed twice this week. Hold six workers after three failed polls instead of relying on stale data.
events-workerfloor 6
After3 failed polls
Operation health
Rolling 60 minutes
2 timeouts · 0 errors
Today’s schedules
All times in Asia/Jerusalem
08:52
Morning checkout peak
checkout-api · 12 → 20
17:45
Evening batch window
events-worker · 18 → 36
23:30
Render queue cooldown
pdf-renderer · 8 → 0
Suggested roadmap
Small features with clear ownership
Each addition keeps the controller focused instead of growing a generic trigger platform.
Now
Ownership guard
Detect HPA or KEDA before writes and create a transfer plan instead of competing for replicas.
Next
SQS failure policy
Choose hold-current or a safe replica floor when queue metrics become stale or unavailable.
Next
Readiness SLO
Measure API acceptance, scheduling, and Ready pods separately so latency claims stay honest.