Resolver did not find any stack that would satisfy requirements and stack characteristics given the time allocated

The resolution process failed as resolver did not find any software stack that would satisfy user requests.

Issue description

The resolution process failed as there was no stack resolved. Check the resolution log to see more details using Thamos CLI:

thamos log

Or directly using the API endpoint:

curl -X GET \
  --header 'Content-Type: application/json' \
  --header 'Accept: application/json' \
  "https://khemenu.thoth-station.ninja/api/v1/advise/python/<adviser-id>/log"

Affected packages

This error message produced is not specific to any package.

Severity

ERROR

Issue fix

First check the resolution log which can give more information on why the resolution process failed. Based on the messages produced, you can address this issue by:

Adjusting requirements

The resolution process might lead to a state when the dependency graph has no valid resolution given version range requirements of the libraries used directly or indirectly (transitive dependencies, dependencies of direct dependencies). In such cases, there is no way to satisfy user requirements for the application stack.

Wait for Thoth to ingest data for unknown packages

If dependencies used in the user stack are not known to Thoth’s knowledge base, Thoth needs to ingest data to respond to user requests. This task is done in the background, transparently to the user. An appropriate message is printed to logs and the resolution process fails.

Adjusting the runtime environment used

Some dependencies might not work for the runtime environment used. An example can be using tensorflow-gpu in version >=1.13.1<=1.15 in a runtime environment running CUDA in version 9. The given TensorFlow releases are not supported when running CUDA 9 hance the resolution process will not resolve any tensorflow-gpu that would satisfy requirements. The solution is to use a TensorFlow release that supports CUDA 9 if the application uses TensorFlow specific features and cannot be updated to a newer TensorFlow release. Another reason can be a package that fails to install to the given runtime environment used (e.g. missing native dependencies, missing manylinux support by pip in the given runtime environment or any other similar issue).

Adjusting the recommendation type used

Some recommendation types, such as stable, might be too strict with respect to packages that can occur in the application stack. Changing the recommendation type might help resolving the software stack, considering implications by using a different recommendation type.

An example could be a “secure” recommendation type request that fails with the error message described. Pipeline units in the secure recommendation type discarded packages that should not be used based on the static source code analysis by Thoth scanners. No “secure” software stack could be produced and hence switching to another recommendation type (e.g. testing) might result in a successful resolution.

See this document that describes recommendation types.

Adjusting the deployment configuration (service provider)

Resolver was not able to resolve any software stack in the CPU time allocated.

See “Configuring and setting up adviser in a cluster” for more info.

Pipeline units

This message is not specific to any pipeline unit.

Recommendation types

This message is not specific to any recommendation type but the recommendation type might affect the resolution process.