diff options
| -rw-r--r-- | f3s/ARGO-ROLLOUTS-SUMMARY.md | 10 | ||||
| -rw-r--r-- | f3s/README-ROLLOUTS.md | 53 |
2 files changed, 30 insertions, 33 deletions
diff --git a/f3s/ARGO-ROLLOUTS-SUMMARY.md b/f3s/ARGO-ROLLOUTS-SUMMARY.md index 2c0372e..80adc23 100644 --- a/f3s/ARGO-ROLLOUTS-SUMMARY.md +++ b/f3s/ARGO-ROLLOUTS-SUMMARY.md @@ -23,10 +23,10 @@ Also registered in ArgoCD: `/home/paul/git/conf/f3s/argocd-apps/cicd/argo-rollou **Replaces**: `frontend-deployment.yaml` (kept for reference) -**Strategy**: Canary with 2-minute observation window +**Strategy**: Canary with 1-minute observation window ``` -Step 1: 50% traffic to new version -Step 2: Pause 2 minutes (observation period) +Step 1: 33% traffic to new version (1 new pod, 3 old pods) +Step 2: Pause 1 minute (observation period) Step 3: 100% traffic to new version (auto-promote) ``` @@ -101,12 +101,12 @@ Added sync option: `RespectIgnoreDifferences=true` to gracefully handle migratio │ │ tracing-demo-frontend Rollout │ │ │ │ ┌──────────────┐ ┌──────────────┐│ │ │ │ │ Stable RS │ │ Canary RS ││ │ -│ │ │ 2 replicas │ │ 1-2 replicas ││ │ +│ │ │ 3 replicas │ │ 1 replica ││ │ │ │ └──────────────┘ └──────────────┘│ │ │ │ │ │ │ │ Endpoints: frontend-service │ │ │ │ - Selects both RS (proportional) │ │ -│ │ - Routes traffic to 50%/100% │ │ +│ │ - Routes traffic to 67%/33% │ │ │ └────────────────────────────────────┘ │ │ │ │ ┌──────────────────┐ │ diff --git a/f3s/README-ROLLOUTS.md b/f3s/README-ROLLOUTS.md index 60ec9b6..b038bf9 100644 --- a/f3s/README-ROLLOUTS.md +++ b/f3s/README-ROLLOUTS.md @@ -23,15 +23,16 @@ just rollout-watch # 5. Trigger in another terminal (Terminal 2) kubectl patch rollout tracing-demo-frontend -n services \ --type='json' \ - -p='[{"op":"replace","path":"/spec/template/spec/containers/0/image","value":"registry.lan.buetow.org:30001/tracing-demo-frontend:latest"}]' + -p='[{"op":"add","path":"/spec/template/spec/containers/0/env/-","value":{"name":"ROLLOUT_V","value":"'$(date +%s)'"}}]' -# 6. Watch progress in Terminal 1 (~4 minutes total) +# 6. Watch progress in Terminal 1 (~90 seconds total) ``` Expected flow: -- 0-2 min: **50% traffic** to new version (canary phase 1) -- 2-4 min: **Wait** for confirmation (canary phase 2) -- 4+ min: **100% traffic** to new version (auto-promoted) +- 0-15 sec: **33% traffic** to canary (1 new pod, 3 old pods) +- 15-60 sec: **Monitor** (paused, observing canary health) +- 60+ sec: **Auto-promote to 100%** (scales all 3 pods to new version) +- ~90 sec: **Complete** (all 3 pods running new version) ## Files Created @@ -61,10 +62,10 @@ Expected flow: - No way to validate before 100% rollout **New way (Rollout with Canary)**: -- 2 old pods → 1 old + 1 new (50/50 traffic) -- Observe for 2 minutes -- If healthy → automatically promote to 2 new pods -- If unhealthy → abort, revert to 2 old pods +- 3 old pods → 3 old + 1 new (33% traffic to canary) +- Observe for 1 minute +- If healthy → automatically promote all 3 pods to new version +- If unhealthy → abort, revert to 3 old pods - Zero downtime, validated before full rollout ## Common Commands @@ -81,10 +82,7 @@ just rollout-status # Detailed info just rollout-info -# Skip waiting, promote now -just rollout-promote - -# Abort and rollback +# Abort and rollback (prevents auto-promotion) just rollout-abort # View history @@ -96,11 +94,11 @@ just load-test ## What Happens During Canary -### Step 1: 50% Traffic (0-2 minutes) +### Step 1: 33% Traffic (0-15 seconds) ``` Frontend Service -├── Stable ReplicaSet (old version): 1 pod → receives 50% traffic -└── Canary ReplicaSet (new version): 1 pod → receives 50% traffic +├── Stable ReplicaSet (old version): 3 pods → receives 67% traffic +└── Canary ReplicaSet (new version): 1 pod → receives 33% traffic ``` Monitor during this phase: @@ -109,18 +107,17 @@ Monitor during this phase: - Logs and traces - Prometheus metrics -### Step 2: Pause (2 minutes) +### Step 2: Pause (15-60 seconds) ``` -Service pauses traffic shift, waiting for: -- Manual promotion via: kubectl argo rollouts promote ... -- Auto-promotion after 2 minutes -- Or abort: kubectl argo rollouts abort ... +Service pauses traffic shift, monitoring canary health: +- Auto-promotion after 1 minute if healthy +- Or abort: kubectl argo rollouts abort ... to stop ``` -### Step 3: 100% Traffic (4+ minutes) +### Step 3: 100% Traffic (60+ seconds) ``` Frontend Service -├── Stable ReplicaSet (new version): 2 pods → receives 100% traffic +├── Stable ReplicaSet (new version): 3 pods → receives 100% traffic └── Canary ReplicaSet (old version): 0 pods → terminated ``` @@ -138,11 +135,11 @@ Updates Rollout resource Argo Rollouts Controller ↓ ├─→ Scales Canary ReplicaSet (1 new pod) - ├─→ Frontend Service routes 50/50 traffic - ├─→ Monitors health/metrics for 2 minutes - └─→ Auto-promotes or waits for manual action - ├─→ If healthy: Scale to 2 new, remove old - └─→ If abort: Remove canary, keep old + ├─→ Frontend Service routes 33/67 traffic + ├─→ Monitors health/metrics for 1 minute + └─→ Auto-promotes if healthy + ├─→ If healthy: Scale to 3 new, remove old + └─→ If abort: Remove canary, keep 3 old ``` ## Demo Scenarios |