Original Ticket:

The SnapshotTooOld be returned from individual shards. This task adds a
StaleSnapshotError error class to error_codes.err that includes ["StaleConfig", "StaleShardVersion", "StaleEpoch", "SnapshotUnavailable", "SnapshotTooOld"] errors
the purpose or this error class is to restart targeting if the snapshot is unavailable or shard version is stale.

As an alternative approach use only ["SnapshotUnavailable", "SnapshotTooOld"] - that will allow to make implementation different handling for SnapshotErrors.

Summary:

Add a new error codes and a new error class to catch snapshot errors that require retrying establishing a global snapshot. This includes: ["SnapshotTooOld", "SnapshotUnavailable"]. "SnapshotTooOld" has already been added as part of Local Snapshot Reads, so only "SnapshotUnavailable" needs to be added.

Details:

In src/mongo/base/error_codes.err:

error_code("SnapshotTooOld", <number>)
error_code("SnapshotUnavailable", <number>)

...

error_class("SnapshotTargetingError", ["SnapshotTooOld", "SnapshotUnavailable"])

POC: https://mongodbcr.appspot.com/184550001