Checkmk Conference #6 goes digital. Get your tickets here!
Updating Checkmk is a bit different than most other software packages that you may be familiar with. Why is that?
The reason is that Checkmk not only allows multiple independent instances (sites) to run on a single server, but it also allows multiple software versions to be installed simultaneously. With this system each instance is assigned to an installed version of the software. To illustrate, we can take the following situation on a fictional server:
Here the instance mysite1 uses version 1.2.8p11.cee, but the instances mysite2 and mysite3 use version 1.2.6p10.cre. The Checkmk-Version 1.2.6p10.demo is actually installed, but is currently not being used.
This example makes it clear that an update does not simply mean the installation of a new Checkmk RPM/DEB package on a server. Another additional step is also required. Let us take the following situation as an example:
Here the instance mysite is to be updated to the Checkmk-version 1.2.8p11.cee. The first step is to download and install the appropriate RPM/DEB-Packet. This is performed exactly as with the initial installation. At first the newly-installed version will not be used by any instance, and it will look like this:
The second step will now be an update of the instance from 1.2.6p10.cre to 1.2.8p11.cee. This is achieved with the omd update command, which we will discuss in detail below:
Following the update, the (likely) no longer required version 1.2.6p10.cre can be deleted by uninstalling the appropriate package. It is however not a bad idea to leave it installed, so that if necessary, with a renewed omd update you can quickly revert to the old version.
2. Updating of Checkmk in detail
2.1. Installation of newer versions
As described in the introduction, the first step with an update is the installation of the desired new version of Checkmk. This is achieved in exactly the same way as with the initial installation – it will however proceed somewhat more quickly since most of the dependent packages have already been installed. In the following example we are installing the package for Ubuntu 14.04 (Codename Trusty):
root@linux# gdebi omd-1.2.8p11.cee_0.trusty_amd64.deb
A list of your installed Checkmk-versions can be displayed at any time with the omd versions command:
root@linux# omd versions 2016.12.18.cee 1.2.8p11.cee 1.2.6p10.cre (default)
One of these listed versions is marked with (default). This default version will be used automatically when creating new sites, as long as no other version is specified with omd -V myversion create mysite. When working with existing sites this is irrelevant. The current default version can be queried with omd version, and it can be altered with omd setversion:
root@linux# omd version 1.2.6p10.cre root@linux# omd setversion 1.2.8p11.cee root@linux# omd version 1.2.8p11.cee
The default version plays no role when updating or managing existing instances. The omd command always starts itself automatically in the version appropriate to the instance.
A listing of the current instances and the versions they use is provided by the omd sites command:
root@linux# omd sites SITE VERSION mysite 1.2.6p10.cre test 2016.12.18.cee
2.2. Performing the update
Once the desired new version has been installed, the instance can be updated. No root-permissions are required for this. The best way to do this is as an instance user:
root@linux# su - mysite
Ensure that the instance has been stopped:
OMD[mysite]:~$ omd stop
The update – in effect switching to a different version – can now simply be performed with the omd update command:
OMD[mysite]:~$ omd update
If more than one target version is available, a selection list will open:
An important part of an update is the refreshing of the originally provided configuration files. Here changes that had possibly been made to these files by the user will not simply be discarded, rather they will be merged. This functions very much like version control systems which attempt to amalgamate changes made to a single file simultaneously by multiple developers.
Occasionally – when the changes affect the same location in the file – that won't function, and a conflict occurs. How you can solve such conflicts will be explained later below.
The update provides a listing of all modified files and directories:
2016-10-11 18:27:07 - Updating site 'mysite' from version 1.2.6p10.cre to 1.2.8p11.cee... * Unwanted var/log/nagios.log (unchanged, deleted by you) * Updated etc/nagvis/nagvis.ini.php * Updated etc/mk-livestatus/nagios.cfg * Updated etc/check_mk/defaults * Updated etc/apache/conf.d/02_fcgid.conf Finished update.
Once everything has been successfully processed, the instance can be switched to the new version...
OMD[mysite]:~$ omd version 1.2.8p11.cee
... and can then be started:
OMD[mysite]:~$ omd start
2.3. Incompatible changes
Software development of course consists of changes. Because we are always actively working to keep Checkmk modern, sometimes cutting dead weight and making changes that turn out to be incompatible is unavoidable. That means that when updating it may possibly be necessary to adapt your configuration, or you should at least check it.
A typical example of such a situation is with new check plug-ins which replace existing plug-ins. If you use one of the affected plug-ins, a fresh service discovery will be required on the affected host.
An overview of all changes in Checkmk, including a search function, can be found online: here. Even more practical however is the built-in search function in the version release notes. This is accessed by clicking on the version number at the top-left of the side bar:
Checkmk tracks new changes automatically, and issues appropriate warnings if they are incompatible:
You can then inspect these ‘Werks’ (‘works’), and approve them with a mouse click. You can also find a listing covering the complete history of the changes, including a search function:
2.4. Downdate – falling-back to an old version
The process of switching back to an older version runs just like an update. To be precise, omd update does not care whether the target version is newer or older than the current version – thus you can switch ‘forwards’ or ‘backwards’ as desired.
Be aware however, that even if a downdate to an older version functions wonderfully, an older Checkmk may not always be able to process data from newer versions. A new Checkmk version may possibly store data and configurations in an extended format that an older version of the software may not understand.
Configurations which are managed in WATO may possibly be converted into a new format once WATO is in active use and can then store the configurations. As long as such actions have not yet occurred, a switch back to an earlier version is generally unproblematic.
Should you be uncertain whether it is necessary to fall back to an earlier version, we recommend to:
- Perform a data backup BEFORE the update.
- Take the time to test the new version before making alterations via WATO.
2.5. The update in detail
Are you curious about what exactly is happening ‘under the hood’ of an update? Or have data conflicts appeared when omd update is running? If so, here is some further reading.
Three actions take place during omd update:
- The refreshing of the default files under etc/ and var/ – i.e., files created by omd create.
- The switching of the active version to the target version by changing the symbolic link version which is found in the Site-directory.
- Post-processing by various packages (e.g., Checkmk). In particular, an Activate Changes will be automatically executed in order to generate a valid configuration for the core.
Actualising files, merging changes
The first step is by far the most comprehensive. Here Checkmk demonstrates a big advantage in comparison to the typical software installation – Checkmk helps you to adapt all of the standard configuration files to the prerequisites of the new version. This resembles the procedure for updating a Linux-Distribution, but goes further in the implementation. Checkmk can handle a multiplicity of situations, for example:
- The merging of file changes with changes made locally by the user
- Files, directories and symbolic links which are obsolete in the new version, or which have been deleted by the user
- Changes to permissions
- Changes to a file type (a symbolic link derived from a file or directory, or vice versa)
- Changes to the target of a symbolic link
Merging and conflicts
If the new version intends changing a configuration file on which the user has also made changes, Checkmk automatically attempts to merge both sets of changes. This is achieved using the same methods as used by version-control systems.
The fewest problems are experienced when your and Checkmk’s changes have a clear physical separation in the text (at least a few lines apart). The merge will then be effected automatically, and without needing the user’s intervention.
If two changes ‘collide’ because they both affect the same location in the data, Checkmk cannot and will not decide which of the changes is more important. In such a situation the user will be alerted, and be able to solve the conflict interactively:
In the situation shown above, you now have the following options:
|d||This shows the differences between the new default version and your version of the file in the form of a ‘unified diff’ (diff -u).|
|y||This is similar to the above, but based on the preceeding default version shows which changes you have made to the file.|
|n||This third option in effect ‘closes the triangle’ by showing the changes which Checkmk intends making to the file.|
|t||By selecting t, your original file – without the already successfully-merged changes – will be opened in an editor. Now edit the file in order to bypass possible conflicts. Once the editor has been closed Checkmk will reattempt the merge.|
|k||Here you can decide whether to accept the data ‘as is’. The successfully inserted changes are retained. Apart from this the file remains as customised by the user.|
|r||With this you can fall back to the old version of your file, and go without Checkmk’s update for this file. Any customisations that may be required must be performed manually.|
|i||Install the new default file: your changes in the old file will be lost.|
|s||If you are uncertain, you can open a shell with s. You will find yourself in a directory containing the relevant file, and there can get a picture of the situation. Quit the shell with Strg-D in order to proceed with the update.|
|a||Abort the update. The instance retains the old version. Files that have already been changed during the update however remain changed! A new update attempt can be started at any time.|
Further conflict situations
Alongside the content-merging of files there is a whole series of further situations in which Checkmk requires your decisions. Some of these are very unusual situations, that nevertheless need to be handled correctly. In these cases Checkmk will always give you the choice of keeping your version, or of adopting the new default version. What is more, there is always the option of aborting an update, or of opening a shell. Examples of such ‘difficult’ situations are:
- conflicting changes to file types (e.g., when a file is replaced by a symbolic link)
- conflicting changes to file permissions
- changed files that are not required by the new sofware version
- files, directories or links created by a user, which conflict with a new version’s files/directories/links
Explanation of the tasks in an update
The update procedure will always output a line of explanation when it makes a change to a file automatically. The following situations are possible – files are referred to here, but this also applies analogously to links and directories:
|Updated||A file has been changed with the new version. Since you have not made a change to the file, Checkmk simply installs the new default version of the file.|
|Merged||A file has been changed with the new version, and at the same time the user has made other changes to the file. Both versions of the file can be merged into one without conflict.|
|Identical||A file has been changed in the new version, and at the same time the user has already made identical changes to the file. Checkmk must not perform any action.|
|Installed||The new version includes a new configuration file which has just now been installed.|
|Identical new||The new version includes a file, an identical copy of which the user has already installed.|
|Obsolete||The new version has obsoleted a file (also applies to a link or a directory). The user has anyway already deleted it. No action.|
|Vanished||Another file is obsolete in the new Checkmk, and the user has neither deleted nor changed the existing version. Checkmk deletes this file automatically.|
|Unwanted||The user has deleted a file which is normally present. Because the version in the new Checkmk has no changes from the last version of the file, Checkmk allows the file to be absent.|
|Missing||The user has already deleted a file, but in the new Checkmk this file contains changes from the previous version. Checkmk installs the new file, and logs a notification of this action to the user.|
|Permissions||Checkmk has updated a file’s permissions because different permissions are set in the new version.|
2.6. Updating without user interaction
Would you like to automate Checkmk’s software updates? You may at first have difficulties with the interactive responses from omd update. There is a simple solution for this scenario: the command has options that have been especially conceived for use in scripts:
- The options -f or --force directly following omd inhibit all types of “Are you sure... ?” questions.
- The option --conflict= directly following update determines the desired behaviour if a file conflict occurs.
Possible values for --conflict= are:
|--conflict=keepold||In the case of a conflict, the user’s own modified version of the file is retained. It is however possible that Checkmk may not be executable, and that manual rectification will be required.|
|--conflict=install||In the event of a conflict, the new standard version of the file will be installed. With this, local changes to the file will be at least partly lost.|
|--conflict=abort||In the event of a conflict the update is stopped. That does not necessarily mean that everything will fall back to the old state. A number of configuration files may have already been updated. The software version will however remain the old version.|
|--conflict=ask||This is the standard procedure, so in this form the option is actually superfluous.|
Below is an example of the complete command for an automated update to version 1.2.8p11 of the mysite instance:
root@linux# omd stop mysite ; omd -f update --conflict=install mysite 1.2.8p11 && omd start
Through the && before omd start a restarting of the instance will be prevented if the omd update is aborted by an error. Replace the && with a semicolon (;) if a start should definitely be attempted even in such a situation.
If you are certain that only a single Checkmk-instance is running on the server, the name to be used in a shell script can simply be trapped in a variable:
root@linux# omd sites --bare mysite root@linux# SITENAME=$(omd sites --bare) root@linux# echo $SITENAME mysite
This enables the above line to be independent of the instance’s name. For example, a small shell script could look like this:
!/bin/bash SITE=$(omd sites --bare) VERSION=1.2.8p11 omd stop $SITE omd -f update --conflict=install $SITE $VERSION && omd start $SITE
3. Upgrading from the demo to the full version
Was your first installation of Checkmk the Free Edition? Once you have a Standard Edition or Managed Services Edition subscription, you can simply upgrade your existing instances to the full version.
The procedure is exactly the same as that for a ‘normal’ update. The only difference is that a version's name with the .demo suffix is upgraded to a name with the .cee suffix. Simply install the desired package of the full version, and switch the existing instance to this with omd update.
This upgrade can be most easily performed if both versions are identical, apart from the .demo and .cee suffixes respectively. What this means for the functionality is that the demo version is completely identical to the full version. Thus an upgrade makes no difference at all.
3.1. Upgrading the Checkmk Appliance
You can also upgrade a demo appliance to a full version with one of the Enterprise Editions without data loss:
- Install a current firmware of the full version on the appliance via its web GUI.
- Install a full version of one of the Enterprise Editions in the version management of the appliance.
- Change the desired instances to that version in the instance management of the appliance.
4. Upgrading from the Raw Edition to the Enterprise Editions
An upgrade of the
Since a number of modules and features of the Enterprise Editions are not available in the Raw Edition, following a changeover there are a couple of points to be aware of. The key point is that when creating new instances of Raw Edition or Enterprise Editions different default settings are set.
Nagios vs. CMC
Since the CRE only supports Nagios as its core, this is preinstalled in instances created by the CRE. This is retained when an upgrade to the CEE is made. That means that after an upgrade, processing will initially continue with a Nagios core. A migration to the CMC is performed with omd config, and this precedure will be described in its own article.
The CEE supports an alternative format for saving historic performance data, one which requires significantly less hard drive-I/O. This is preinstalled in new CEE-instances. CRE-instances will not be changed over automatically by an upgrade. How the migration can be performed is described in its own chapter in the article covering Performance data and graphing.
The CRE has no notification spooler. Thus following the changover to the CEE it is not active at first. How to activate it can be learned here.
5. Uninstalling Checkmk
The uninstallation of no longer required Checkmk versions is performed using the operating system’s package manager. To do this, enter the installed package’s name – NOT the file name of the original RPM/DEB file. Important: Only delete Checkmk versions that are no longer being used by any instance!
Checkmk instances that are no longer required can simply be removed with omd rm (thereby deleting all data as well!):
root@linux# omd rm mysite
SLES, RedHat, CentOS
Here is how to identify which Checkmk packages are being used in RPM-based systems:
root@linux# rpm -qa | grep check-mk check-mk-enterprise-2016.05.19.0 check-mk-enterprise-2016.10.11.0 check-mk-raw-1.2.8b9.0 check-mk-raw-1.2.8p11.0
The deletion is performed with rpm -e:
root@linux# rpm -e check-mk-raw-1.2.8b9.0
Use the below to identify which packets are installed:
root@linux# dpkg -l | grep check-mk ii check-mk-enterprise-2016.05.19 0.trusty amd64 Check_MK is a full featured system monitoring ii check-mk-enterprise-2016.10.11 0.trusty amd64 Check_MK is a full featured system monitoring ii check-mk-raw-1.2.8b9 0.trusty amd64 Check_MK is a full featured system monitoring ii check-mk-raw-1.2.8p11 0.trusty amd64 Check_MK is a full featured system monitoring
The uninstallation is performed with dpkg --purge:
root@linux# dpkg --purge check-mk-raw-1.2.8b9 (Read database ... 505719 Files and directories are currently installed.) Remove from check-mk-raw-1.2.8b9 (0.trusty) ... Delete the configuration files from check-mk-raw-1.2.8b9 (0.trusty) ...
6. Files and directories
The files and directories relevant to this article can be found here. As always, paths that do not begin with ‘/’ apply after the home directory of the (/omd/sites/mysite) instance:
|version||Symbolic link to the installation of the Checkmk version used by this instance.|
|/omd/versions||Within this directory a subdirectory exists for every installed Checkmk version. The files belonging to root and are never changed.|
|/omd/sites||Within this directory, for every instance there is a home directory containing its configuration files and variable data. This data belongs to the instance’s user, and can be changed by configuration and operations.|
|/usr/bin/omd||Management command for Checkmk instances. This is a symbolic link to the default version’s bin-directory. When a particular instance is accessed the omd-command substitutes itself with that of the appropriate version.|