CoreWeave Production Checklist
Community-contributed. Not affiliated with, endorsed by, or sponsored by CoreWeave, Inc. CoreWeave is a registered trademark of CoreWeave, Inc.
Overview
Use this gate before exposing a CoreWeave inference or training workload to
production. It proves that capacity, data protection, observability, and recovery
are owned and tested; it is not a substitute for a service-specific risk review.
Prerequisites
- A signed deployment change, named service owner, and approved production namespace.
- Passing staging evaluation, capacity forecast, and rollback manifest.
- On-call routing, secret-manager references, and data classification for the workload.
Instructions
- Complete each checklist item with a link to its evidence, not a verbal assertion.
- Confirm a staged rollout, health check, and rollback command during the change window.
- Stop the release if any security, data, capacity, or observability gate lacks an owner.
- Record the final go/no-go decision, time, owner, and redacted verification receipt.
Inference Services
Storage
Security
Monitoring
Rollback
kubectl rollout undo deployment/my-inference
kubectl rollout status deployment/my-inference
Output
- A production-readiness record linking capacity, storage, security, monitoring, and
rollback evidence to the deployment.
- A clear go/no-go decision with a responsible owner and tested recovery command.
Error Handling
| Failed gate |
Required action |
| Readiness or staging test fails |
Hold the release and restore the prior known-good revision. |
| GPU capacity is unconfirmed |
Do not launch; resolve quota or capacity with the platform owner. |
| Secret or network policy is missing |
Block deployment until the namespace hardening baseline is present. |
| Monitoring route is absent |
Block production exposure until alerts reach the on-call owner. |
Examples
Run the last production checks during the approved change window:
kubectl -n inference-prod apply -f release.yaml
kubectl -n inference-prod rollout status deployment/my-inference --timeout=15m
kubectl -n inference-prod get pods -l app=my-inference
If rollout or the smoke test fails, execute the documented rollback immediately,
capture redacted events, and mark the change no-go rather than retrying with weaker
security or capacity controls.
Resources
Next Steps
For upgrades, see coreweave-upgrade-migration.
1---2name: coreweave-prod-checklist3description: Production readiness checklist for CoreWeave GPU workloads. Use when launching inference services, preparing GPU training for production, or validating deployment configurations. Trigger with phrases like "coreweave production", "coreweave go-live", "coreweave checklist", "coreweave launch".4license: MIT5---6# CoreWeave Production Checklist
7
8> **Community-contributed.** Not affiliated with, endorsed by, or sponsored by CoreWeave, Inc. CoreWeave is a registered trademark of CoreWeave, Inc.
9
10## Overview
11
12Use this gate before exposing a CoreWeave inference or training workload to
13production. It proves that capacity, data protection, observability, and recovery
14are owned and tested; it is not a substitute for a service-specific risk review.
15
16## Prerequisites
17
18- A signed deployment change, named service owner, and approved production namespace.
19- Passing staging evaluation, capacity forecast, and rollback manifest.
20- On-call routing, secret-manager references, and data classification for the workload.
21
22## Instructions
23
241. Complete each checklist item with a link to its evidence, not a verbal assertion.
252. Confirm a staged rollout, health check, and rollback command during the change window.
263. Stop the release if any security, data, capacity, or observability gate lacks an owner.
274. Record the final go/no-go decision, time, owner, and redacted verification receipt.
28
29## Inference Services
30
31- [ ] GPU type and count validated for model size
32- [ ] Autoscaling configured (KServe or HPA)
33- [ ] Health and readiness probes set
34- [ ] Resource requests AND limits specified
35- [ ] Node affinity targeting correct GPU class
36- [ ] `minReplicas >= 1` for production (no cold starts)
37
38## Storage
39
40- [ ] Model weights in PVC (not downloaded at startup)
41- [ ] Checkpoints saved to persistent storage
42- [ ] Storage class appropriate (SSD for inference, HDD for archival)
43
44## Security
45
46- [ ] Secrets for model tokens and registry access
47- [ ] Network policies applied
48- [ ] Container images from trusted registries
49
50## Monitoring
51
52- [ ] GPU utilization metrics collected
53- [ ] Inference latency and throughput tracked
54- [ ] Alert on pod restarts and OOM events
55- [ ] Log aggregation configured
56
57## Rollback
58
59```bash
60kubectl rollout undo deployment/my-inference
61kubectl rollout status deployment/my-inference
62```
63
64## Output
65
66- A production-readiness record linking capacity, storage, security, monitoring, and
67 rollback evidence to the deployment.
68- A clear go/no-go decision with a responsible owner and tested recovery command.
69
70## Error Handling
71
72| Failed gate | Required action |
73|---|---|
74| Readiness or staging test fails | Hold the release and restore the prior known-good revision. |
75| GPU capacity is unconfirmed | Do not launch; resolve quota or capacity with the platform owner. |
76| Secret or network policy is missing | Block deployment until the namespace hardening baseline is present. |
77| Monitoring route is absent | Block production exposure until alerts reach the on-call owner. |
78
79## Examples
80
81Run the last production checks during the approved change window:
82
83```bash
84kubectl -n inference-prod apply -f release.yaml
85kubectl -n inference-prod rollout status deployment/my-inference --timeout=15m
86kubectl -n inference-prod get pods -l app=my-inference
87```
88
89If rollout or the smoke test fails, execute the documented rollback immediately,
90capture redacted events, and mark the change no-go rather than retrying with weaker
91security or capacity controls.
92
93## Resources
94
95- [CoreWeave CKS](https://docs.coreweave.com/docs/products/cks)
96
97## Next Steps
98
99For upgrades, see `coreweave-upgrade-migration`.
gabrielmoreira