Good evening from a wet colorado fall night. I want to say first, thank you so much to the ISC team, you have provided an amazing array of top class software and I have enjoyed using it. Thank you.
Ok, let me get straight to it. I have a been steadily upgrading through the various 9.x versions, and most recently have had 9.5.0-P1 running, and have followed all the posts on this list with respect to the memory leaks, etc. I, too, have noticed this. What I did was put an entry in my crontab file (yes, I'm old school, and use cron on windows to time my jobs...) to restart bind every 8 hours. This has been an ok workaround...the memory gets freed, and bind starts up clean. I loose the cache, to be sure, but at least I don't run out of pagefile. I got the note that 9.5.0-P2-W1 was out, and upgraded. I just do the upgrade manually, by replacing files by hand, its just easier that way and I know where everything goes. I do have to reboot to free the lock on bindevt.dll but other than that, its straightforward. All seemed to be going fine, then I noticed bind wasn't resolving after an hour or so. It responds fine to an rndc status, but not to dig or nslookup; it just times out. In addition, the log file shows the following entry: client: error: UDP client handler shutting down due to fatal receive error: end of file (I'm aware of this on 'nix, but this is win32 mind you...) I also noted in the win event app log, the following upon startup: the 'files' limit (2048) is less than FD_SETSIZE (16384), increase 'files' in named.conf or recompile with a smaller FD_SETSIZE. I don't use the files option...which should default to unlimited, so what that means is a mystery. I put in a files option with 16385 and still got the error notice. I've rolled back to 9.5.0-P1 with my cron 8 hour restart for now. So...any ideas guys? cheers!