Index reload

Index reload is periodical background process synchronizing in-memory matcher indices with storage.

If an index disappears from storage, index is dropped from the matcher’s memory as well.

If a new index with a new label appears in storage, the matcher will try to load the new index into memory.

If a new index appears in the storage with a newer version of the label than the index loaded into the memory of the matcher, it will try to load the new index into memory instead of the old one. To ensure a specific index can be reloaded by only one matcher at a time, the redlock mechanism is used. If the lock is acquired, the older version of the index is dropped from the matcher’s memory, and the newer one is loaded.

If a problem occurs while loading the index, e.g. insufficient memory, a message is sent to the log & monitoring.