StoreFailedException

A write failed against the filesystem, and the store has stopped accepting new ones.

The reason it stops rather than carries on: a failed append may already have written part of a record. Appending the next commit behind those bytes would leave a log whose readable prefix ends at the failure, so every commit after it — all of them acknowledged — would be sitting behind bytes recovery cannot get past. Refusing further writes keeps the damage to the commit that actually failed.

Reads keep working: the memtable is untouched by an IO fault, and a caller that has just lost its writer usually needs to read the data out. DocumentStore.close also still works, and reopening recovers the prefix that was written before the fault.

The cause is the original filesystem failure, and it is the thing to look at.