Skip to content

Change GKE node pool machine type to e2-standard-2 - #1

Open
canvascloudai wants to merge 1 commit into
mainfrom
test/cwm-gke-e2-standard-2
Open

canvascloudai wants to merge 1 commit into
mainfrom
test/cwm-gke-e2-standard-2

Conversation

@canvascloudai

Copy link
Copy Markdown
Owner

Changes google_container_node_pool.primary_nodes machine_type from n1-standard-1 to e2-standard-2.

@cloud-world-model

cloud-world-model Bot commented Oct 26, 2026 •

Copy link
Copy Markdown

☁️ Cloud World Model — Infrastructure Simulation Report

Deterministic 30-step comparison: dab7ea0 → 5e9e161 · 2026-10-26T17:18:42.236Z

Merge review summary

No unfavorable changes detected in comparable modeled metrics; modeled cost up $0.0688/hr (+$50.22/month at 730 h) (trade-off).

Evidence is incomplete or limited; no overall pass is inferred.

Modeled improvements

  • P99 latency: 373 → 277 ms
  • Peak aggregate CPU (maximum across simulation steps): 93.5 → 66.2%
  • Resilience: 0 → 76 score points
  • Goodput: 282 → 300 RPS
  • Error rate: 6.00 → 0.00%

Cost trade-off

  • $0.7323/hr → $0.8011/hr; +$0.0688/hr (+$50.22/month at 730 h)

Higher modeled cost alone is not an operational regression and may buy capacity or performance.

Evidence limitations

  • Performance evidence is diagnostic or otherwise unvalidated, not a validated production finding.
  • Some resource mappings, connections, service classifications, or shape/count/region values use inference, fallback, or defaults. See Coverage below for the affected entries.
  • Recorded scope warnings apply; inspect the comparison details for the full limitations.
  • Modeled comparison only: no live infrastructure, production traffic, application-code or provider validation. No rollback simulation or merge-safety guarantee.
Metric Before After Impact
Modeled total cost per hour $0.73/hr (modeled) $0.80/hr (modeled) +$0.0688/hr (+$50.22/month at 730 h) · cost trade-off
p99 latency 373ms (uncalibrated reference estimate) 277ms (uncalibrated reference estimate) -96ms · favorable
Peak aggregate CPU (maximum across steps) 93.5% (reference estimate) 66.2% (reference estimate) -27.3 percentage points · favorable
Offered client traffic 300.0 RPS 300.0 RPS 0.0 RPS
Goodput 282.0 RPS 300.0 RPS +18.0 RPS
Error rate 6.00% 0.00% -6.00 percentage points
Resilience 0/100 (modeled) 76/100 (modeled) +76 score points · favorable

Offered client RPS is the shared denominator for goodput and error rate. Goodput is already net of errors and modeled shedding; modeled shed traffic is included in errors and is not counted twice.

Diagnostic scope: Generic fallback performance outputs and warning/error events are diagnostic model estimates, not validated critical or saturation findings. Owned typical predictions require the matching workload/topology through the 300 RPS holdout; matching workloads from 301–500 RPS use unvalidated diagnostic extrapolation, while unmatched workloads remain generic. Modeled numbers and scope warnings are retained; threshold indicators and impact directions are illustrative, not production validation.

Metric evidence: Cost and resilience are modeled estimates, not measured bills or outcomes. Peak aggregate CPU is the maximum reported aggregate CPU metric across simulation steps; CPU events can report a separate individual resource reading.

  • Before: app weight typical (defaulted); owned calibration fit not applied. Load scope: not calibrated.
  • After: app weight typical (defaulted); owned calibration fit not applied. Load scope: not calibrated.

Traffic by simulation step — Before (base) and After (PR/head)

Step Before offered Before goodput Before error rate After offered After goodput After error rate
1 300.0 RPS 282.0 RPS 5.85% 300.0 RPS 300.0 RPS 0.00%
2 300.0 RPS 282.8 RPS 5.75% 300.0 RPS 300.0 RPS 0.00%
3 300.0 RPS 281.0 RPS 6.20% 300.0 RPS 300.0 RPS 0.00%
4 300.0 RPS 280.0 RPS 6.60% 300.0 RPS 300.0 RPS 0.00%
5 300.0 RPS 283.6 RPS 5.45% 300.0 RPS 300.0 RPS 0.00%
6 300.0 RPS 283.5 RPS 5.50% 300.0 RPS 300.0 RPS 0.00%
7 300.0 RPS 283.5 RPS 5.50% 300.0 RPS 300.0 RPS 0.00%
8 300.0 RPS 281.0 RPS 6.30% 300.0 RPS 300.0 RPS 0.00%
9 300.0 RPS 280.0 RPS 6.55% 300.0 RPS 300.0 RPS 0.00%
10 300.0 RPS 283.0 RPS 5.65% 300.0 RPS 300.0 RPS 0.00%
11 300.0 RPS 281.0 RPS 6.25% 300.0 RPS 300.0 RPS 0.00%
12 300.0 RPS 280.6 RPS 6.45% 300.0 RPS 300.0 RPS 0.00%
13 300.0 RPS 283.8 RPS 5.40% 300.0 RPS 300.0 RPS 0.00%
14 300.0 RPS 283.6 RPS 5.45% 300.0 RPS 300.0 RPS 0.00%
15 300.0 RPS 283.8 RPS 5.40% 300.0 RPS 300.0 RPS 0.00%
16 300.0 RPS 282.0 RPS 6.00% 300.0 RPS 300.0 RPS 0.00%
17 300.0 RPS 282.0 RPS 5.95% 300.0 RPS 300.0 RPS 0.00%
18 300.0 RPS 283.8 RPS 5.40% 300.0 RPS 300.0 RPS 0.00%
19 300.0 RPS 282.0 RPS 6.00% 300.0 RPS 300.0 RPS 0.00%
20 300.0 RPS 281.0 RPS 6.25% 300.0 RPS 300.0 RPS 0.00%
21 300.0 RPS 283.8 RPS 5.40% 300.0 RPS 300.0 RPS 0.00%
22 300.0 RPS 283.0 RPS 5.65% 300.0 RPS 300.0 RPS 0.00%
23 300.0 RPS 281.6 RPS 6.15% 300.0 RPS 300.0 RPS 0.00%
24 300.0 RPS 283.0 RPS 5.60% 300.0 RPS 300.0 RPS 0.00%
25 300.0 RPS 281.0 RPS 6.30% 300.0 RPS 300.0 RPS 0.00%
26 300.0 RPS 280.9 RPS 6.35% 300.0 RPS 300.0 RPS 0.00%
27 300.0 RPS 279.8 RPS 6.75% 300.0 RPS 300.0 RPS 0.00%
28 300.0 RPS 280.9 RPS 6.35% 300.0 RPS 300.0 RPS 0.00%
29 300.0 RPS 282.0 RPS 5.95% 300.0 RPS 300.0 RPS 0.00%
30 300.0 RPS 282.0 RPS 6.00% 300.0 RPS 300.0 RPS 0.00%

Cost impact

Modeled total cost per hour: $0.73/hr before → $0.80/hr after (+$0.0688/hr (+$50.22/month at 730 h)).
This modeled total includes direct resource charges and simulation-derived effects; it is not a real-world bill.

Direct modeled resource charges (expanded Terraform rows grouped only when their source and modeled pricing match):

  • Before:
  • primary_nodes: $0.195/hr
  • After:
  • primary_nodes: $0.234/hr

Simulation-derived effects (not direct source-resource charges):

  • Egress/data transfer: $0.537/hr before → $0.567/hr after (+$0.030/hr).
    The egress delta is an emergent, non-additive result of the simulated topology and workload; it is not attributable to an individual source change.

Cost reconciliation: Before $0.195/hr direct + $0.537/hr derived = $0.732/hr; After $0.234/hr direct + $0.567/hr derived = $0.801/hr.
Snapshot monthly resource charges are compute/resource-only; the modeled total above includes derived effects.

Simulation basis

Evaluated Terraform root: ./
Modeled resources within ./: 1 before → 1 after
Workload: synthetic 300 RPS · Budget: 30 steps

Repository scope: 1 touched IaC file plus 1 required support file at the base ref; 1 touched IaC file plus 1 required support file at the head ref. Live infrastructure, application code, and production traffic were not evaluated.

Simulation inputs, mappings and assumptions

Observed

  • Before (terraform): 1 touched IaC file: gke.tf; 1 fetched support file: terraform.tfvars.
  • After (terraform): 1 touched IaC file: gke.tf; 1 fetched support file: terraform.tfvars.

Mapping outcomes

Exact describes source construct identity: the provider and CWM resource type were preserved. Inferred describes CWM assumptions added for modeled behavior, sizing, defaults, or topology. A source construct can be Exact at the identity level and still have separate inferred modeling characteristics; this does not change its classification, simulation results, telemetry outcomes, or pricing behavior.

Exact

  • No exact source mappings were recorded.

Inferred

  • Before: primary_nodes — Inferred GCP GKE node pool.
  • After: primary_nodes — Inferred GCP GKE node pool.

Unsupported

  • No recognized source constructs were left unmodeled.

Assumed

  • Each modeled side used the same synthetic 300 RPS workload for 30 simulation steps.

Named source-to-model mappings

Inferred — source capacity model

  • Before: primary_nodes — GCP GKE node pool; n1-standard-1 × 2 workers; $0.04749975/node-hr × 2 = $0.095000/hr worker cost (us-central1 reference). Modeled capacity: 1500 RPS/node × 2 = 3000 RPS, not observed or guaranteed. Cluster management fee is separate. Source region: us-central1.
  • After: primary_nodes — GCP GKE node pool; e2-standard-2 × 2 workers; $0.06701142/node-hr × 2 = $0.134023/hr worker cost (us-central1 reference). Modeled capacity: 3000 RPS/node × 2 = 6000 RPS, not observed or guaranteed. Cluster management fee is separate. Source region: us-central1.

Not evaluated

  • Live infrastructure
  • Application code
  • Production traffic

Coverage

  • Before primary_nodes — inference from google_container_node_pool → modeled shape n1-standard-1.
  • After primary_nodes — inference from google_container_node_pool → modeled shape e2-standard-2.

Base: 41dd05db-a67a-4017-a283-fcb6f65f355c · View base snapshot · Head: 96208a3d-f673-4da0-8406-a197a2f8ff9d · View head snapshot

Scope warnings

  • ⚠️ Generic fallback performance outputs and warning/error events are diagnostic model estimates, not validated critical or saturation findings. Owned typical predictions require the matching workload/topology through the 300 RPS holdout; matching workloads from 301–500 RPS use unvalidated diagnostic extrapolation, while unmatched workloads remain generic. Modeled numbers are retained; no production measurements were evaluated.

Simulation diagnostic events

Warning/error severities describe modeled engine conditions, not validated critical or saturation findings. Calibration of a metric does not validate these events.

Metric scope: Before Peak CPU comes from the Compute fleet aggregate CPU metric (maximum across simulation steps). A resource-specific event (for example, main nearing capacity) describes that resource, not the Before aggregate.

  • ⚠️ Warning · Before: · Resource: Compute fleet Diagnostic model event (not validated): Compute fleet CPU exceeded 80% threshold (maximum 91.7%)

Jump to the full ordered event log →

🔋 10 credits consumed from your CWM workspace.

IaC detected: terraform

Files analysed (2)

gke.tf
terraform.tfvars


Powered by Cloud World Model · Manage settings

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant