Using restart instead of reload causes lots of problems. 1) when the server is slow to stop, it will not come up again 2) when the certificate private keys have a password, restart will fail (not wanting unencrypted private keys in backup is a valid reason to use passwords, IMHO) 3) even if everything works well, the server is unresponsive for 10 seconds On the other hand, using reload might cause problems with some third party modules, as Adam pointed out in #298689. Unfortunately, I can't find references which modules are affected. So, how to proceed from here? 1) and 3) might be mitigated by using graceful-shutdown instead of stop, but this has not been tested. I think the cleaner solution is to switch to reload during logrotate and see what breaks. Then those other modules should be fixed or removed (or the problem documented). Now (i.e. early in the lenny release cycle) is the right time to make this change in unstable. For stable, one can wait and see how this works out in unstable, or try to use graceful-shutdown. What do you think? Should I make this change now?
Attachment:
signature.asc
Description: This is a digitally signed message part.