If bind_port (UDP) or start_port (TCP) is used, restarted members might reuse the same port, therefore one could run into reincarnation issues.
Solution: if enabled, we maintain a file in the temp dir (shared by all instances on the same host), into which we write ports chosen. When a new socket is created, we read the file, weed out expired ports and pick the next port. Then we write that back to the file and close the file again.
- relates to
-
JGRP-130 Problems with reincarnation
- Resolved