Bug 8131

Summary:	ntpd: should retry on name resolving error
Product:	Busybox	Reporter:	Floris Bos <bos>
Component:	Networking	Assignee:	unassigned
Status:	RESOLVED WONTFIX
Severity:	enhancement	CC:	busybox-cvs
Priority:	P5
Version:	1.23.x
Target Milestone:	---
Hardware:	PC
OS:	Linux
Host:		Target:
Build:

Description Floris Bos 2015-05-18 19:48:11 UTC

Currently ntpd tries to resolve the hostnames in /etc/ntp.conf on startup, and quits instantly with a "ntpd: bad address '0.pool.ntp.org'" error if any of them  do not resolve.

I think it would be better if it tried again later, instead of quiting.
In some environments network connectivity is not available instantly at boot  (e.g. networks that use STP).

Comment 1 Denys Vlasenko 2016-04-04 17:50:46 UTC

This was discussed on the ML and rejected.
If you want to cope with not-working DNS resolution and resulting failures, babysit your services.

Comment 2 Floris Bos 2016-04-04 18:06:08 UTC

>If you want to cope with not-working DNS resolution and resulting failures, 
>babysit your services.

You are wrongfully assuming here that it is the DNS server that is the problem.

While it is typically the device running busybox instead that does not have working connectivity yet on boot.
On Ethernet switches that have the spanning tree protocol enabled it takes 45 seconds after the network link goes up before it goes to forwarding state and there is a working network connection...

Comment 3 Denys Vlasenko 2016-04-06 00:02:09 UTC

(In reply to Floris Bos from comment #2)
> You are wrongfully assuming here that it is the DNS server that is the problem.
> While it is typically the device running busybox instead that does not have working connectivity yet on boot.

No, I do agree with you, DNS often doesn't work right after boot because network is not yet up.

The solution to this and many other problems is to babysit your services.