Please convert init script to upstart

Bug #604717 reported by Chuck Short on 2010-07-12
This bug affects 7 people
Affects Status Importance Assigned to Milestone
ntp (Ubuntu)
Ante Karamatić

Bug Description

Binary package hint: ntp

As per the server-maverick-upstart-conversion specification, please review the debdiff to convert the init script to upstart.

Chuck Short (zulcss) wrote :
Thierry Carrez (ttx) on 2010-07-14
Changed in ntp (Ubuntu):
importance: Undecided → Wishlist
status: New → Confirmed

Should the dh_installinit really be marked "upstart only"? I am thinking more about merging back to Debian or even backporting where upstart may not be in use. Otherwise I think this patch looks fine, and I may try a test build to verify it. I happen to agree ntp should migrate to upstart.

tags: added: patch
Daniel Hahler (blueyed) on 2010-07-15
Changed in ntp (Ubuntu):
status: Confirmed → Triaged
Colin Watson (cjwatson) wrote :

I agree with David's comment.

-invoke-rc.d --quiet ntp start >/dev/null 2>&1 || true
+start stop >/dev/null 2>&1 || true

This is wrong - should be 'start ntp'.

(Incomplete review; I still need to look at the job itself.)

Colin Watson (cjwatson) wrote :

+ test -f /etc/default/ntp && ./etc/default/ntp || true

Missing a space between '.' and '/etc/default/ntp'. This is probably why you had to add the '|| true' - remove that. Also, to match the init script this should be test -r not test -f.

+ test -e /var/lib/ntp.conf.dhcp && NTPD_OPTS="$NTPD_OPTS -c /var/lib/ntp/ntp.conf.dhcp"

The first filename is typoed - should be /var/lib/ntp/ntp.conf.dhcp.

You only seem to have ported part of the locking code over. You need the backgrounded lockfile-touch and the kill as well (otherwise the lock could time out on a very slow start), although quite how you're going to get the pid of the lockfile-touch process across to the post-start script I'm not quite sure.

Please try to avoid the use of a pidfile. Upstart's process tracking should be sufficient.

The apparmor code in the Upstart job appears to be new relative to the init script. Why?

Colin Watson (cjwatson) wrote :

The last point may be skew between what I was reviewing and current Maverick - a number of packages have been converted to dh_apparmor. I haven't looked into exactly what's involved here.

Chuck Short (zulcss) wrote :

Ill fix up all of these mistakes and resubmit the debdiff, I did test
them but I guess I didnt hit these problems while testng them.


EVERYTHING YOU NEED for a great and exciting day you can buy HERE!
Only TODAY ANY GOODS you can get with a next code 5486 which gives you a 40% dis pw count.
Be prepared for a scary day and make a show!

Get the great di fe scou rtg nts on popular so em ft loo wa jbs re today at
All s rf of hiz wa qup re is instantly available to do hpq wnl jsf oad - No Need Wait!
USA, English, France, Italy, Spanish, German and more!!!

SO je FTW wus ARE:Windows 7 Ultimate 32 bit99.95Windows 7 Ultimate 64 bit99.95Windows XP Professional with Service Pack 369.95Office Professional Plus 2010 32-bit89.95Office Professional Plus 2010 64-bit89.95Adobe Photoshop CS5.1 Extended99.95Office Professional 200769.95Adobe Acrobat 9 Pro Extended59.95Office Home and Student 200749.95
Also we have so mu up ch s yrf of tkd t for MA gh CIN fi TO bwj SH!!!Adobe Creative Suite 5.5 Master Collection for MAC269.95Adobe Creative Suite 5.5 Design Premium for MAC219.95Microsoft Office 2008 Standart Edition for MAC119.95Aperture 3 for MAC79.95Adobe Photoshop CS5.1 Extended for MAC89.95
To re rny vi zw ew full list of the offers, v vfl is jx it

isoma (isoma) wrote :

The ntp package provides /usr/sbin/ntp-wait

NTP can use this, or a similar technique, to emit an Upstart event when the time is synchronised. It would be nice to be able to react to this (although maybe not to block daemon startup indefinitely).

Ante Karamatić (ivoks) wrote :

This debdiff adds an upstart job for ntp. Upstart job does the same thing SysV script does. In addition, it adds two features that are common requirement in production:

1) does one off sync before starting ntpd in foreground - off by default
2) syncs hardware clock after the one off sync - off by default

Many services, OpenStack's nova comes to mind, require time to be synced, rather ntpd started. By doing one off sync before starting ntpd, system will have correct time before nova-* services start. Nova service then needs to depend on started ntpd.

Hardware clock sync might be an overkill, but, IMHO, doing it right after we synced the OS time seems the best place to do it.

Ante Karamatić (ivoks) wrote :

Found few typos. Fixed.

Changed in ntp (Ubuntu):
assignee: nobody → Ante Karamatić (ivoks)

I know it's a long time, but I'm cleaning up old NTP bugs atm.

These days handled via systemd init script wrapper.
IMHO I don't think it still makes sense to add/convert upstart jobs for ntp.

If anything a native systemd job, but then systemd takes care of most basic time sync tasks on its own anyway.
That said - closing the bug by setting Opinion (which it is - an "old" opinion - in todays point of view)

Changed in ntp (Ubuntu):
status: Triaged → Opinion
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Duplicates of this bug

Other bug subscribers