NTILZA & all
AFAIK this change/improvement is a design decision by developers, so that a reconfiguration and reboot only needs to be advised/done when absolutely necessary.
There are thousands of sme servers being supported by sme support professionals, and this change minimizes the workload and reduces disruption to end uses, when doing upgrades.
So unless specifically requested, there is no need to do both a post-upgrade and reboot, although for purposes of simplicity there is a need to do a reboot.
Further to the above, there is still a requirement to manually restart specific services or do a reboot after specific packages are updated.
Some package updates require no action (neither reconfiguration or reboot).
It is really a case of "it depends on what gets updated/installed". See bug report quoted comments below.
If e-smith* or smeserver* packages are updated, the system will request a reconfiguration (signal-event post-upgrade) and reboot.
So, it really depends if you understand Linux and know what you are doing when updating, or not:
In some cases you can do nothing.
In some cases you need to restart only specific services.
In some cases you need to reboot.
In some cases (and you will be notified) you need to do both reconfiguration and reboot (signal-event post-upgrade; signal-event reboot).
Therefore any user who does not understand when to apply the above criteria, needs to at least do a reboot after a set of updates is applied.
If you receive advice to do a reconfiguration and reboot, then you must do both (signal-event post-upgrade; signal-event reboot).
The above applies to both command line updates and server manager updates, as both methods effectively do the same thing.
To quote from bug 5709
http://bugs.contribs.org/show_bug.cgi?id=5709> Please confirm: these updates are exempt from the signal-event
> post-upgrade/reboot requirement if updated via the command line (or web
> console)? That the application of these updates do NOT require a reboot of the
> server or the additional commands??
That will always depend on the exact set of updates. Many updates require no
additional action (e.g. strace updated). Some will require restart of specific
services (e.g. if php or apache is updated). Some will require a reboot (e.g.
new kernel). None will require a full reconfiguration - only the e-smith-* and
smeserver-* packages might require full reconfiguration.
> Is there now a notification where a services-specific
> restart is required (like php or apache),
No.
> or is this (was this always?) done
> during the YUM update (service is shutdown, updated, then restarted)??
No.
> If not,
> then the average user would just end up performing a restart of the system ?
Correct.
> It will be clearer to users doing yum updates if at the end of the procedure,
> yum advises the user if the usual signal-event post-upgrade / reboot is needed,
> and if not, explicitly states that no further action is required. I am assuming
> also that services affected are restarted automatically by yum.
Such an NFR already exists. However, it is a complex problem in AI to examine a
set of update packages and determine the minimal set of actions to activate
them all. Thus far all that has been implemented (and in reality all that is
very likely to be implemented) is to take the safety first action of doing a
full reconfigure and restart.
5 Replies
1295 Views