Readiness check
Before publishing a server, open Platforms in its card and run the readiness check for your marketplace. Fix the reported problems, then run it again. The check helps prepare your server; the marketplace makes its own verification decision.
Check results
Section titled “Check results”| Result | What to do |
|---|---|
| Pass | the requirement is met |
| Warn | read the explanation and check whether the warning affects your setup |
| Fail | fix the problem before publishing |
| Unknown | check that the server is online, repeat the check, and complete any requested manual confirmation |
Use the requirements and supported versions shown in Platforms. They may change, so check them again after hardware, driver, or network changes.
Server preparation
Section titled “Server preparation”- System and driver: use a supported OS and a working NVIDIA driver. Schedule driver and system updates between rentals.
- GPU: check that all cards are detected, temperatures are available, and there are no hardware warnings.
- Containers: Docker must be running, with NVIDIA Container Toolkit providing GPU access to containers.
- Network: check the public IP address, connection speed, and inbound ports. If your connection uses CGNAT, ask your provider about a public address.
- Marketplace agent: install it using the marketplace’s instructions. Platforms shows whether it runs.
If a sensor or hardware check fails, follow the instructions in that row. Missing readings do not always mean a hardware fault.
Vast.ai requirements
Section titled “Vast.ai requirements”Check GPU and CUDA compatibility, memory, storage, network speed, and ports. Some requirements depend on the number of GPUs. Run the check again after adding or replacing a card.
Docker storage and quotas
Section titled “Docker storage and quotas”Follow the storage requirements in the check, including the filesystem and container disk quotas. If the quota test fails, use the troubleshooting guide.
If Vast agent or Machine identity fails, install the Vast agent with the host setup on Vast.ai, wait for registration, and select Re-check now.
Clore.ai requirements
Section titled “Clore.ai requirements”Check the supported GPU models, OS and driver versions, storage, and network. Pay attention to warnings about overclocking or a changed power limit, including limits set in Flotera’s Tuning tab. Follow Clore’s requirements before qualification.
QuickPod requirements
Section titled “QuickPod requirements”Flotera checks QuickPod installation and machine identity on
Ubuntu 22.04 GPU hosts. Platforms reports the daemon version, a fingerprint
of its machine UUID, and the configured port range. The fingerprint is not an
API key. Linking requires a connected QuickPod account and a fresh probe whose
fingerprint matches a machine in that account.
The card also shows the last reported states of quickpod.service and
quickpod-hostblind.service, with the age of that report. A stale “running”
report does not prove the services are running now.
QuickPod requires a public IPv4 address and an inbound port range forwarded to
the host. Its own listing and port verification are separate steps; an installed
daemon does not mean the machine is available for rentals. Connection, listing
controls, pricing strategies and earnings are described in the
QuickPod guide. hostblind shows “no unit” on a machine without GPU
passthrough; that is not an error.
See QuickPod’s machine setup
and hosting overview.
If a check still fails
Section titled “If a check still fails”Open readiness check failures. If you need support, include the check name, expected and actual values, and the error message.