unmerge 432322 thanks Philipp Kern wrote: > severity 432322 serious > thanks > > On Sun, Dec 30, 2007 at 12:01:15PM +0100, Philipp Kern wrote: >> On Mon, Jul 09, 2007 at 01:40:31PM +0100, Jon Dowland wrote: >>> The last two times that the network-manager package has >>> been upgraded, the process of doing so has killed my active >>> network connections, including the ssh session with the >>> controlling tty for the aptitude upgrade process: >> That's so freakingly annoying. Couldn't you simple restart the daemon >> through invoke-rc.d in postinst instead of stopping it in preinst >> and bringing it back up in postinst? What's the reason for this >> madness? > > The Release Managers consider this bug to be RC for Lenny. Would it be > possible to reload NetworkManager in a way not causing the network > connection to break, i.e. without the devices being deconfigured?
Hi Philipp et al, my first idea to address this issue, was to modify/patch NM to not deconfigure the device when NM is stopped and make it possible for NM to pick up the existing connection when NM is started again (that's also the reason why I merged 432322 and 433087, as it would have solved both issues). I talked about this with upstream, and he said for connections like ppp/vpn/wireless it would be almost impossible, for ethernet connections maybe but also only with considerable effort. > If all the files are replaceable while NetworkManager is still running > `dh_installinit --no-start --no-restart-on-upgrade' together with > invoke-rc.d in postinst (as n-m has an init script now, not just an > event script) would be a start as it would reduce the downtime, but > it will still deconfigure the interfaces shortly and take some time > to reaquire the IP (on LAN) and on WiFi more time to reassociate to > the AP, possibly needing user intervention. I will take this approach for now: restart NM in postinst (instead of stop in prerm, start in postinst). I'll do the same for dbus. This should minimise the downtime and in case the upgrade process get's interrupted in between keep the machine accessible via the network. It's not a perfect solution but I hope RMs will consider it appropriate. > And is #433087 (network-manager: Behaves poorly when dbus daemon is > restarted) really the same issue? Dear maintainers, please provide > some input on this issue. As said, I first wanted to address this differently. For now I unmerged the bugs again and will handle them separately. Cheers, Michael -- Why is it that all of the instruments seeking intelligent life in the universe are pointed away from Earth?
signature.asc
Description: OpenPGP digital signature