[[openSUSE] openSUSE 10.3-373267] Dovecot requires ntp which requires named operational

Open Source Support Tools

Home

Just Starting? View the Demo

Download Diagnostic Tools

(JavaScript Disabled)

Search Item

Summary

Reported Issue

Title: [[openSUSE] openSUSE 10.3-373267] Dovecot requires ntp which requires named operational

Project: opensuse

Item Last Modified: Mon, 29 Sep 2008 04:59:00

Tags:

Tag This

Bookmark This

Details

[[openSUSE] openSUSE 10.3-373267] Dovecot requires ntp which requires named operational

Reporter:		Brian Harris brian
Created:		Mon, 24 Mar 2008 06:20:00
Updated:		Mon, 29 Sep 2008 04:59:00
Key:		373267
Versions:		Not provided
Environment:
Priority:		5
Status:		Resolved
Resolution:		FIXED
Original Link:		https://bugzilla.novell.com/show_bug.cgi?id=373267
Summary:		Dovecot requires ntp which requires named operational

Description:

The basic problem was that dovecot was aborting a few seconds after startup
because ntp set the system time back 6 seconds after the system had been turned
off a couple of days.
There are two issues:
Firstly /etc/init.d/dovecot needs ntp added to its 'Required Start' list.
Secondly /etc/init.d/ntp needs the named daemon up and running. Although $named
is in the /etc/init.d/ntp 'Required Start' list, the ntpdate command gets
issued before named is operational, causing ntpdate to fail. A simple
work-around is to sleep 3 seconds before issuing the ntpdate command.

Comments:

Cristian Rodriguez	Mon, 24 Mar 2008 08:02:00
> Firstly /etc/init.d/dovecot needs ntp added to its 'Required Start' list. nope. it needs $time in Should-Start. > Secondly /etc/init.d/ntp needs the named daemon up and running. Although $named > is in the /etc/init.d/ntp 'Required Start' list, the ntpdate command gets > issued before named is operational, causing ntpdate to fail. A simple > work-around is to sleep 3 seconds before issuing the ntpdate command. That seems to be a different bug.
Marcus Rueckert	Mon, 24 Mar 2008 15:26:00
have you ever considered fixing the hwclock so you dont get 6s jumps on every boot?
Brian Harris	Tue, 25 Mar 2008 10:50:00
The 6 seconds discrepancy was after 60 hours, which is not abnormal for PC clocks in my experience. I have since discovered the OpenSuse article on SuSE Package Conventions and see that $named should indicate that name resolution is active. There may have been a network glitch causing the ntpdate to fail. I will remove the 3 second delay and monitor the situation over the next week or so and double check that it consistently fails at boot.
Brian Harris	Tue, 08 Apr 2008 04:16:00
I can confirm that without the 3 second delay, ntpdate always fails.
Diederik van der Boor	Mon, 05 May 2008 01:19:00
This bug has a low severity, but I hope this could be bumped up. According to http://wiki.dovecot.org/TimeMovedBackwards running ntpd/ntpdate after dovecot is started could cause dovecot to die immediately. Not something you'd like to see happening when rebooting your server. Note the docs also recommends running ntpd instead of ntpdate from cron. In some systems ntpd/ntpdate is run at boot, but only after Dovecot has started. That can cause Dovecot to die immediately. If you have this problem, fix your init scripts to run ntpd/ntpdate before Dovecot.
Marcus Rueckert	Mon, 29 Sep 2008 04:59:00
applied the same fix for 10.2/10.3 as part of the security fix

URL: ../../opensuse/issues/373267 • Email URL