Mailinglist Archive: opensuse (878 mails)

< Previous Next >
Re: [opensuse] what to do after zypper dup
lynn said the following on 09/14/2013 09:11 AM:
12.3
Hi
I have some repos which need updating so I run zypper dup. If I run
zypper up, they don't get updated.

Personally I don't think you should be doing that.
Do you really understand the difference between 'up' and 'dup'?

When there is a update that I want/need which 'up' isn't giving me I can bring it in using

zypper install package.revision

If you simple try

zypper install package

it will tell you that a later revision exists and the full command to install it.

That you see items in 'dup' that aren't in 'up' might simply be that you have settings on the priority levels of distributions set to make it come out that way. In this case it becomes an all or nothing.




After the show, I get:
There are some running programs that use files deleted by recent
upgrade. You may wish to restart some of them. Run 'zypper ps' to list
these programs.
and then:
You may wish to restart these processes.

This makes perfect sense.

First, recall that Linux does not delete a file that is being help open. We often use this trick to autodelete temporary files. q.v.

If there is a program already running that uses a library at revision level 8 and suddenly it has to use the same library with revision level 9 then unpredictable things are going to happen. BAD! So we don't let that happen.

This is much clearer if you think in term of something like editing a text file with SED. If the file suddenly changes then the edits you made are no longer to make any sense. What was line 132 is now line 166 and changing what is now 132 becomes a mistake.

The application already has the library/8 or its own binary 'open' and continues to use that. If the program exits (or is killed) and is restarted then it uses the new binary of library.

Please note that it says "you *MAY* want to restart these processes".
That *MAY* is important. You may be quite happy running the older versions for now. The changes in the new version may not be critical.

That answers the "what if I can't reboot" case for a lot of instances - you don't need to restart the processes.

In many more instances you may only need to log out and in again to start over all the user GUI applications, including the X server.

What's left? Systems process and the kernel.

Restarting the kernel needs a reboot just as installing a new kernel as part of installing a new distribution needs a reboot. No way around that. The issue is "how critical?" If something is failing then the users are demanding it being fixed so a reboot is necessary -- what they are doing isn't working anyway. If its not critical then waiting for scheduled PM -- you do schedule PM, don't you? - is the way to go.

Heck, isn't this what 'scheduled PM' amounts to anyway?

Other system processes are easier to restart with systemd, I've found, but again its an issue of scheduling. Paying attention to when things like cron jobs get started; email flow; ... it all depends on what you are running. Maybe you're not a full load 7/24 so you can come in at 4am to restart the sshd. Otherwise it has to be scheduled.

If you are a 7/24 system then maybe you should be running HA or hot-failover so that one machine can be updated then cut over for any further new jobs and let the ones on the other expire. Certainly with a connectionless protocol like HTTP(s) having multiple servers and load balancing for a 7/24 is good practice and would facilitate updates that happen as a matter of course.


Say there are 10 processes. Is there something like:
zypper restart-processes-after-dup?

I most certainly hope not!

Normally I can reboot or dup when we have fewer users, but what if I can't?

The simple answer is "it may not matter'
The longer answer is that "it if does matter that much then the users will be wanting it so they will put up with logging off for the reboot".
Unless you have things messed up then its not as if a reboot is going to take long. My slowest server, the one that has to restart a huge DNS table, normally reboots in around 2 minutes. If I simply pull the plug so it has to run the FSCKs (in parallel of course) that might stretch to 5 minutes.

That's less time that it takes to to take the lift down to the 2nd floor, go to the cafeteria, get a coffee and come back.


YMMV. One place I worked there was a coffee hutch just down the corridor from my cubicle but the machine took about a minute to make a fresh cup.




--
How long did the whining go on when KDE2 went on KDE3?

The only universal constant is change. If a species can not adapt it
goes extinct. That's the law of the universe, adapt or die.
-- Billie Walsh, May 18 2013
--
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse+owner@xxxxxxxxxxxx

< Previous Next >
Follow Ups
References