Tahoe-LAFS storage server retains incident logs indefinitely

Created by: exarkun

When Tahoe-LAFS detects an "incident" it dumps a bunch of log data to its incidents directory.

These logs may contain somewhat arbitrary information available to the storage server because they can contain Python tracebacks which are generally indiscriminate about what they report on. This may include such values as storage indexes, ZKAPs, and client IP addresses.

Over time the logs may accumulate in such number that significant storage is dedicated to them. They may also degrade server performance since they are all collected in one directory and directory performance often suffers when the number of contained files gets large enough.

Also, in principle, an incident represents a flaw in the software. If this is true, the incidents should probably be drawn to our attention (perhaps via some kind of monitoring system). If not, maybe we should upstream some work to Tahoe-LAFS to eliminate them (at least the useless ones).