Configure the OpenTelemetry Collector filestorage extension to persist the sending queue across restarts for durable delivery

Steps

1. Add a file_storage extension block under extensions, setting directory to a path the Collector process can read/write (default /var/lib/otelcol/file_storage on non-Windows, %ProgramData%\Otelcol\FileStorage on Windows); set create_directory: true if it may not already exist.
2. Reference the extension's component ID inside the target exporter's sending_queue block, e.g. sending_queue: { storage: file_storage }, so the queue persists to a bbolt-backed write-ahead store instead of memory only.
3. Add file_storage to the top-level service.extensions list — defining the extension alone does nothing until it is both listed there and referenced by an exporter's sending_queue.storage.
4. Tune durability/perf as needed: timeout (default 1s) for file-lock waits, fsync: true to force an fsync after every write (safer but slower), and max_size to cap each bbolt database file's size in bytes (0 = unlimited).
5. Optionally configure compaction (compaction.on_start, compaction.on_rebound with rebound_needed_threshold_mib / rebound_trigger_threshold_mib / check_interval) to reclaim disk space after the queue drains following a prolonged outage.