TroubleshootingOfficial documentation11 min readIssue landscape

FastGPT Deployment and Environment Issue Landscape

Explore Deployment and Environment Issue Landscape with symptom-based checks, published article links and practical guidance for troubleshooting your deployment.

This page groups the 721 published deployment and environment documents by the stage at which the problem appears. Use it to narrow down the stage first, then open the list for that stage to find the specific document. Grouping follows the technical objects named in document titles.

Narrowing down: three steps

StepWhat to confirmPoints to
1Whether the service process started, and the state of container and orchestration componentsProcess not up → containers and orchestration, images and architecture; process up but unreachable → startup and reachability
2Whether backend logs show connection errors from dependencies (database, object storage, inference service)Database or storage errors → databases and object storage; model call errors → model serving and inference
3Whether the problem appeared after a version or configuration changeAfter a change → upgrades and migrations, environment and configuration; unrelated to a change → follow the conclusion from steps 1 and 2

Stages, how to tell them apart, and document counts

StageHow to tellPublished documentsList
Images and architecturecontainer image retrieval and processor architecture13Search Technical Center
Containers and orchestrationcontainer runtime and orchestration18Search Technical Center
Databases and object storagedatabase connections and object storage17Search Technical Center
Upgrades and migrationsthe upgrade process and behaviour after upgrading114Open list
Environment and configurationenvironment variables and configuration files125Open list
Startup and reachabilityservice startup and reaching the service13Search Technical Center
Model serving and inferenceself-hosted inference services and model integration41Open list
Agent and MCPagent runtime, sandbox and MCP tooling29Search Technical Center
API and authenticationAPI calls and authentication33Open list
Workflow and nodesworkflow orchestration and node configuration37Open list

The five stages with lists cover 350 documents. Another 281 documents cover more scattered topics and are not yet grouped into a stage; they can be found by searching for the specific error text.

Three ordering mistakes to avoid

  1. Changing configuration before reading logs. Most deployment problems name the component and the error code in the backend log. Adjusting environment variables or orchestration files before reading the log removes the baseline for everything that follows.
  2. Going straight to the main service and skipping its dependencies. When the database, object storage or inference service is unreachable, the main service looks broken in every case, so starting there leads nowhere.
  3. Not separating before and after a change. A problem that appeared after an upgrade or configuration change follows a different path from one that appeared during steady operation; confirming the time relation halves the search space.

Checks before go-live and before any change

  1. Confirm every environment variable added or removed by the target version has been handled
  2. Confirm each companion component image matches the main service version
  3. Confirm database and object storage connection details work in the target environment
  4. Confirm the processor architecture matches the selected image
  5. Confirm the inference service is reachable from inside the deployment network, and complete one call to verify

What this landscape does not cover

Grouping follows the technical objects named in document titles. A document may touch several stages; it is placed in the first stage it matches. The following are outside the scope of this page:

  • Deployment forms and configuration specific to the commercial edition
  • Network and permission setup coupled to a specific cloud provider
  • Performance and capacity questions that need runtime data to judge

Keep reading

References

If the problem is still not located

The entries above cover cases that can be reproduced from public information. If the problem depends on configuration details of a specific deployment, or needs runtime logs to confirm, contact sales for deployment-stage support; the cloud service can be used directly without handling environment dependencies.

  • Contact sales: support for self-hosting and upgrades
  • Get started: use the cloud service and skip environment setup
  • Pricing: compare what the cloud and self-hosted forms cover