[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Checking for services to be restarted on a default Debian installation

On 7 September 2014 15:30:22 CEST, "David Prévot" <taffit@debian.org> wrote:
>Le 07/09/2014 02:07, Paul Wise a écrit :
>> On Tue, Sep 2, 2014 at 2:48 AM, Thijs Kinkhorst wrote:
>>> My questions to this list:
>>> - Do people agree that this would be something that's good to have
>in a
>>> default installation? Are there drawbacks?
>Not restarting by default the DM seems to be nice thing to have.
>How does it work if the upgrade run in the background? Will all needed
>service be restarted without asking? (If so, the gdm3 restart issue may
>be a blocker).

As a long time user and system administrator I agree that notification and *optional* automatic restarts have a place in the default install (with appropriate notes in the changelog for Jessie, obviously!).

For a server, there should be some easy to adjust setting, choosing between automatic restarts and simply notifying of "restart of x, y, z needed due to upgrade b and c (with comment from changelog: is this a security issue?)".

Do we have a framework for "persistent" gui notifications on the desktop? Eg: next time someone in the sudo group logs in; show request for system restart/kexec and/or subsystem restarts? I know Ubuntu has a default software center thing for that -- is there something like it in tasksel-desktop? (I generally run a lean xmonad-only setup - a notification in my xmobar would be nice, though)

On a server I'm generally happy with an email to root - but do we have somewhere we could put notifications? Eg: service names in /var/run/restart-pending or something along those lines?

The idea being that apt/dpgk/checkrestart could append package names here, and a "do-pending-restarts"-script could remove them (probably better just to run checkrestarts again and verify start time/loaded libraries vs latest installed version and update the "needs-restart" queue as appropriate?).

The more I think about, the better I like the idea of having a text-file as a job queue of pending restarts, and a script that checks running processes for open dlls that updates such a file (can be put in cron for generatoøing gui alerts w fallback to console alerts on systems w/o xorg).

Alerting for restarts amounts to checking for the presence of such a file and re-running the checkrestart script to regenerate it, or remove it if all needed restarts are done (seperate file for kernel, or use service name kexec? For servers it might nice to notify on updated inintrd/grub.cfg as there is no *guarantee* the system will boot after such changes -- until they've been verified by a successful reboot).

Thoughts? Is this overboard for getting into Jessie?

Best regards, 


Via phone - please excuse quoting and spelling

Reply to: