Log-Out | Dirk Mittler's Blog

Successful X-server Update Today

Today the Debian / Jessie package maintainers pushed through an update to my X-server, bringing that up to version ‘2:1.16.4-1+deb8u1‘. In a case like this, it’s better to restart the session, just to load the new X-server version, but also, to make sure that any problems in running the X-server do not take us by surprise, during some later reboot.

As it happens, the computer this took place on, which I name ‘Phoenix‘, is also my Web-server. And so it’s desirable to perform the restart of the session, without doing a full reboot of the computer. Under Linux, with the ‘KDE’ desktop manager, we can do this by just Logging Out, and then Logging back In. Doing so also causes an X-server restart.

Not doing a full reboot of my computer, also meant that my Apache Web-server continued running, without causing any disruption to the visibility of my Web-site, or of this blog.

Yet, there was a glitch. As I told my user-session to log out, the display hanged, with the mouse-pointer still visible, but with the rest of my display black, and with no further activity on the hard-drive. In this situation I can help myself, by using the KDE, <Ctrl>+<Alt>+Backspace key-combination, to force an X-server restart. Doing so successfully ended my user-session. Then, the X-server restarted, and I got to log back in.

What this means is that the preparatory work I did according to this earlier posting, has paid off.

I count this as a successful update.

Dirk

KDE Update This Morning

This morning, effectively as I slept, my Linux computer named ‘Phoenix’ downloaded and installed a set of updates, to KDE, which is the framework installed on this computer, for managing the display of my Desktop / GUI, as well as for PIM. Because this update installed itself unattended, it meant that user-processes were still running, which did not have the latest version of the software loaded into RAM.

This can happen easily on Linux computer-systems, which generally just keep working that way – in the short term. But in the long term, eventually something is going to want to wake up a KDE-program which is not already running, and which will want to use the latest images, at which point some sort of incompatibility could set in.

Fortunately though, KDE will only affect user-processes, as I just wrote, but will not affect system-processes, such as my Apache Web-server, or the instance of the MySQL-server, which is running as a system- rather than as a user-process. And so what I can do in this case, is just log out the user, and log him back in. Doing so also reloads the X-server, which does a shutdown as well, and restarts the graphics-driver.

Both the update to KDE, as well as my log-out, log-in maneuver, worked well, without any hint at first glance, of anything going wrong. As a result, the availability of the site and my blog were unaffected. Also, the ‘memcached‘ process is a system-process that did not require any restart, so that the responsiveness of my blog, to the most-frequent requests for postings from readers, should still be as quick as it was only 30 minutes ago.

Dirk

User-Session Restart Today

When most of us want to make sure that our laptops are stable, and have saved their user-data, and have a fresh slate of available memory again, we just perform a reboot. However, on a tower computer which is also acting as a Web-server, this may not be such a viable option, because a true reboot would also cause said Web-server to go offline briefly. But as an alternative, we can log out the current user without shutting down the system, and then log back in again. On most Linux systems that have real desktop managers, this also restarts the X-server, and frees some additional graphics memory.

I just performed such a User-Session Restart, on the computer I name ‘‘. When I do this, it actually empties out the large amount of data which my system has swapped out. What tends to happen on this machine, which only possesses 4GB of RAM, is that its heavy KDE Desktop Manager has resource leaks, which continuously cause memory to be allocated but not used. The unused memory gets paged to the swap partition. But because it is also a 64-bit machine, there is no real risk that the system would run out of 64-bit virtual addresses, corresponding to this leaked resource. Only, the used portion of the swap partition becomes embarrassingly large. A log-out followed by a log-in, forces the desktop environment to save all its data, and in so doing, also forces all the data which has been swapped out, to be swapped back in, simultaneously.

I believe that this is reason, why actual restarts on this one machine are really so slow. It takes several minutes just for the wallpaper to disappear, and then the mouse-cursor, and then for an actual X-server-restart to take place.

At the same time, this exercise has confirmed what I already thought, which was that a ‘gstreamer’ update weeks ago, did not affect my sound server adversely. I still have sound.

Also, in the province of Quebec, Daylight Savings Time started several weeks ago, and whenever this happens, my Desktop Notification Icon, ‘‘, starts to fire off user-notifications for scheduled events at an illogical time but persistently, around 23h00 or so, until 24h00. The only way to stop that in the short term, is actually to disable these alarms, until the desktop is given a new session. Then, as experience on my laptop ‘‘ already confirmed, can be re-enabled, and those illogical, wrongly-scheduled alarms stop happening.

Finally, when and if there finally is another power failure, the risk of data-loss becomes lower.

None of this should have affected system services, nor the Web-server, and therefore not my site either.

Dirk

Akonadi Server Update Successful Today

My Linux computers have PIM Software installed (…Personal Information Management), which is implemented with the KDE Desktop Manager, through a service which runs in user-space, called ‘Akonadi’. This service allows personal information to be synchronized across several platforms, including Google, Given the wishes of the user.

The most common way in which Akonadi is installed, is with its MySQL back-end for storing information locally, and doing so enables the software developers to run an instance of the MySQL database server, not as a system service, but by default as a localized instance, that accesses the folders of one user and only requires the user-name of one user.

According to this earlier posting, an update to the MySQL software itself, broke this way in which Akonadi works – probably because Debian developers did not coordinate that update with KDE developers. According to what I had written there, the system-wide package ‘mysql-server‘ needed to be installed, that also launches the system-wide daemon, before the form of it localized for use with Akonadi would run again. This upset the way I had installed my Laptop ‘Klystron’, where I had previously not thought it necessary to install the system-wide server, to enjoy the use of Akonadi.

Just this evening, KDE developers and Debian / Jessie package maintainers pushed through an update to several Akonadi-related packages, which I installed using my package manager. When we do this, user processes such as Akonadi are not restarted automatically, since in doing that, the package-manager would also break the current session.

And so what I needed to do in order to apply the update, was just to log out of my current, graphical desktop session, and log back in again.

But according to this earlier posting, there has been a history as well, of such attempted log-outs hanging, leaving just the mouse-pointer showing, and leading to no further, apparent progress in actually logging out. So, as of October 21, I had enabled the <Ctrl>+<Alt>+Backspace key-combination on all the desktop sessions, which needs to be arranged before the desktop session is started, so that I would effectively be able just ‘to blow away the current session’, in case a log-out hangs again.

This time around, while I was performing the log-out, my procedure did hang again, leaving me waiting for a few minutes, taking from 19h55 until 20h05 in total. But I did use the <Ctrl>+<Alt>+Backspace key-combination, to hasten the restart of the X-server, which meant that I was not forced to reboot the computer.

This turned the update into a Success, because system services such as the Apache Web-server and ‘memcached‘, and the system-wide MySQL database server, did not need to be restarted. And as a result, there was no disruption in any form to the blog or to my site.

Yay!

Now, Because I have the system-providing mysql-server package installed already both on my Laptop, as well as on this server-box ‘Phoenix’, I cannot tell whether this update did in fact fix that issue. Also, I cannot be sure that the update was meant to fix that issue. All I know, is that nothing is broken as it stands, and no services were disrupted.

Dirk