Automatic graphics processing unit selection based on known configuration states

A method comprising:

• executing, on a first node of a computing cluster, a first instance of a computing process that is configured to use a first graphics processing unit (GPU) in a first GPU configuration;
• detecting a loss of functionality that affects the computing process;
• and
• responsive to detecting the loss of functionality, automatically determining that a second instance of the computing process can execute on a second node by:
• determining that the second instance of the computing process can execute using a second GPU in a second GPU configuration on the second node;
• and
• configuring the second instance of the computing process to use the second GPU in the second GPU configuration on the second node,
• wherein the second GPU configuration is different from the first GPU configuration;
• and
• wherein the second GPU configuration is based on at least one known configuration state.

A system comprising:

• a storage medium having stored thereon a sequence of instructions;
• and
• a processor that executes the sequence of instructions to cause the processor to perform acts comprising,
• executing, on a first node of a computing cluster, a first instance of a computing process that is configured to use a first graphics processing unit (GPU) in a first GPU configuration;
• detecting a loss of functionality that affects the computing process;
• and
• responsive to detecting the loss of functionality, automatically determining that a second instance of the computing process can execute on a second node by:
• determining that the second instance of the computing process can execute using a second GPU in a second GPU configuration on the second node;
• and
• configuring the second instance of the computing process to use the second GPU in the second GPU configuration on the second node,
• wherein the second GPU configuration is different from the first GPU configuration;
• and
• wherein the second GPU configuration is based on at least one known configuration state.