If you set $named_wait_host to a remote hostname and something is going wrong with the name resolution process for that particular hostname, your system will hang indefinitely. Some services (like ntpd) depend on being able to resolve remote DNS hostnames, but you would still want the system to come up if it is having trouble resolving $named_wait_host. It would help to have an additional variable that specifies an upper bound for how long /etc/rc.d/named spends attempting to resolve $named_wait_host. That way $named_wait_host can be set to a remote hostname but your system will not be completely dependent on that hostname successfully resolving. This addition would make it easy to solve problems like the one discussed here: http://forums.freebsd.org/showthread.php?t=9171 Fix: I've attached patches for /etc/rc.d/named and /etc/defaults/rc.conf ; these patches add and document an optional $named_wait_max variable that sets the maximum number of seconds that rc.d/named will wait while trying to resolve $named_wait_host. Patch attached with submission follows: How-To-Repeat: Set up rc.conf with the following variables: ntpd_enable="YES" named_enable="YES" named_wait="YES" named_wait_host="someunresolvablehostname.somenonexistentdomain.com" Then run: /etc/rc.d/named restart
I misfiled this; it should have been filed in conf. Sorry about that.
Responsible Changed From-To: freebsd-bugs->freebsd-rc Fix category and assign.
Responsible Changed From-To: freebsd-rc->dougb dougb has volunteered to look at named bugs.
State Changed From-To: open->closed I've thought about your PR for a while, and my conclusion is that the current setting is doing what it's designed to do. If name resolution is critical to the system's function, subverting the _wait option is not what we should be doing. If it isn't critical, the user would not have set that option in the first place (which defaults to off btw). I will keep your PR on record in case this issue resurfaces down the road. Meanwhile, thanks for your interest in making FreeBSD better. Doug
I believe it was an error to reject this patch. A problem exists and needs a fix like this for the reasons stated when the bug was opened. However, I believe the default behavior should be to wait forever unless the user specifically configures some timeout value. I'm reopening this because it's potentially the fix for bug 231193.
Keyword: patch or patch-ready – in lieu of summary line prefix: [patch] * bulk change for the keyword * summary lines may be edited manually (not in bulk). Keyword descriptions and search interface: <https://bugs.freebsd.org/bugzilla/describekeywords.cgi>
With a heavy heart, return this PR to rc@ with bugmeister hat. To Andy: is this aging PR still relevant?
> To Andy: is this aging PR still relevant? It is not still relevant to me, personally. I don't know if it's still a failure mode / bug that exists today and is worth fixing.