Issue

After a storage failure at the filesystem level that necessitates a restart of the A-MQ broker, if the persistence store is KahaDB, the broker will enter a KahaDB recovery state. The broker will not start until recovery is complete. The recovery process will read the journal (.log) files, and try to reconstruct the message index files.

Unfortunately, there are some filesystem corruptions that cannot be recovered completely, and a few that will defeat the recovery procedure completely, and cause it to spin endlessly. Service can be interrupted for hours or, in the worst cases, indefinitely.