Fault isolation using code paths

A non-transitory computer-readable storage that stores instructions which, when executed by one or more processors, causes the one or more processors to perform steps comprising:

• receiving a first request to perform a first operation that may be performed by any one of a plurality of code paths within a server;
• selecting a first code path, of the plurality of code paths, to perform the first operation;
• detecting occurrence of a fault during performance of the first operation by the first code path;
• identifying a first entity involved in the first operation that experienced the fault;
• in response to the first entity being involved in the first operation that experienced the fault, determining that quarantine criteria for the first code path has been satisfied relative to the first entity;
• in response to determining that quarantine criteria for the first code path has been satisfied relative to the first entity, storing data that indicates that the first entity is quarantined relative to the first code path;
• after storing data that indicates that the first entity is quarantined relative to the first code path, receiving a second request to perform a second operation that can be performed by any one of the plurality of code paths;
• in response to the second request, determining whether the first entity is involved in the second operation;
• in response to determining that the first entity is involved in the second operation and that the first entity is quarantined relative to the first code path, selecting a second code path of the plurality of code paths to perform the second operation;
• wherein the second code path is different from the first code path;
• wherein the method is performed by one or more computing devices.

A storage server comprising:

• storage for storing data that belongs to databases managed by one or more database servers;
• a first code path for handling I/O requests from the one or more database servers;
• a second code path for handling I/O requests from the one or more database servers;
• a code path selector configured to determine whether entities involved in I/O requests are quarantined from the first code path and, when an entity involved in a particular I/O request is quarantined from the first code path, causing the particular I/O request to be handled by the second code path;
• a fault handler configured to store information about which entities are involved in faults when faults occur while the first code path is being used to service I/O requests;
• and
• a quarantine handler configured to determine, based on which entities are involved in faults, whether quarantine criteria is satisfied for those entities, and to store data that indicates that the entities are quarantined in response to determining that the quarantine criteria is satisfied.