After a server crash, the server rebooted and sid-filter tried to start. As the unix socket was present on the file system, the service refused to start. In the logs, the following lines appeared : sid-filter[26560]: Sendmail Sender-ID Filter: Unable to bind to port local:/var/run/sid-filter/sid-filter.sock: Address already in use sid-filter[26560]: Sendmail Sender-ID Filter: Unable to create listening socket on conn local:/var/run/sid-filter/sid-filter.sock sid-filter[26560]: smfi_opensocket() failed Abviously the service wasn't running. Reproducible: Always Steps to Reproduce: 1. crash the server, or try "pkill -9 sid-filter" to simulate the crash 2. /etc/init.d/sid-filter restart Actual Results: The service doesn't start. Expected Results: The service has started. As to have a quick fix on this problem, please see attachment to this bug. It's a modification of of the init script. The following test have been done : - normal restart - normal start (stop then start) - start when service is already running - start after service crash and socket present
Created attachment 217439 [details, diff] init script
By the way, as you can see in the patch, the warning message is taken from the mysql init script and is present to warn the admin. I also checked if the process was still running to avoid a 2nd start (just in case). Also in case of an inet socket, the function ignores the checking. All the above tests are fine in both unix and inet sockets.
Hi Gentoo. Is there anyone who wish to comment/commit on this bug ? It's been there for 2 months now with a working solution but nothing happened. thanks
Fixed in rev r5. Thanks and sorry for the long delay.