ID: 13739
Title: jenkins_nodes: Allow negative timespan for Clock difference
Component: Checks & agents
Level: 1
Class: Bug fix
Version: 2.2.0i1
The Clock difference of a Jenkins node can be negative
if the node's time is running ahead.
In this case the check would crash because negative values
cannot be rendered as a timespan.
Now the absolute value of the time is used and displayed.
ID: 13579
Title: Expand discovery of FAST LTA checks
Component: Checks & agents
Level: 1
Class: Bug fix
Version: 2.2.0i1
On some Fast LTA systems, the discovery of the fast_lta_* checks was not triggered.
This is fixed now by using a more open snmp_scan_function.
ID: 13704
Title: apc_symmetra: Transformation of old rulesets failed
Component: Checks & agents
Level: 1
Class: Bug fix
Version: 2.2.0i1
Transformation of old rules faild during <tt>cmk-update-config</tt>.
The displayed warning was "The entry State if calibration is invalid is missing".
ID: 13812
Title: KUBE kube_cpu use latest cpu usage value when no new data is available
Component: Checks & agents
Level: 1
Class: Bug fix
Version: 2.2.0i1
The check returned a 0 cpu usage value when the Kubernetes special
agent was polled too frequently. This werk fixes this behaviour
and returns the last reported usage value for a time interval of
60 seconds if no new data is available from the agent. If the last
reported value is older than 60 seconds than the usage value will
be omitted.
ID: 12703
Title: Drop experimental PySNMP backend
Component: agents
Level: 1
Class: Bug fix
Version: 2.2.0i1
The experimental PySNMP is dropped because it did not perform
as expected. The classic backend will be selected instead for
people who have set PySNMP previously.
ID: 13094
Title: Various ValueSpec-related API cleanups
Component: Checks & agents
Level: 2
Class: New feature
Version: 2.2.0i1
Various changes have been done to <tt>ValueSpec</tt> and its subclasses,
improving maintainability and robustness. If you don't write your own
complicated <tt>ValueSpec</tt> subclasses and you don't have any self-written
complicated check parameters, you are probably not affected.
<ul>
<li>
<tt>ValueSpec</tt>'s constructor doesn't accept any random unspecifed
keyword arguments anymore. It is better to raise an exception at the
incorrect call site than silently accepting every typo. To fix any problems
with this change, just remove the offending keyword arguments, they served
no purpose after all.
</li>
<li>
<tt>ValueSpec.canonical_value()</tt> is an abstract method now, otherwise we
force lots of places to handle <tt>None</tt>. Previously quite a few things
worked only more or less by accident. If you implement your own
<tt>ValueSpec</tt> subclass, you have to implement that method now, unless
an intermediate class already does it for you.
</li>
<li>
<tt>DropdownChoice</tt> and <tt>CascadingDropdown</tt> don't have a
<tt>no_preselect</tt> keyword argument anymore. It was redundant, because
<tt>no_preselect_title</tt> is enough: If this keyword argument is not
<tt>None</tt>, <tt>no_preselect=True</tt> is implied. Previously one could
specify nonsensical combinations of those arguments. To fix any problems
with this change, just drop the <tt>no_preselect</tt> keyword argument.
</li>
<li>
<tt>Alternative</tt>, <tt>UploadOrPasteTextFile</tt>, <tt>TextOrRegExp</tt>,
and <tt>CAorCAChain</tt> don't have a <tt>style</tt> keyword argument
anymore. It didn't serve any purpose since Checkmk 2.0. Just remove it.
</li>
</ul>
ID: 13639
Title: Fix displaying agent output on crash report page
Component: Multisite
Level: 1
Class: Bug fix
Version: 2.2.0i1
When the crash report page displayed a agent based check report, it displayed
an error message "Internal error: Unsupported type <class 'bytes'>" in the
"Agent output" box.
The agent output can now be displayed again, even if it might not be 100%
correct in case there are characters included in the agent output which can not
be encoded in the default page encoding (UTF-8). This is OK, because the agent
output there is only for informational purposes. Users that want to have the
full and correct agent output can download the crash report archive and extract
the agent_output file from it.
ID: 13084
Title: Reintroduce housekeeping job before activation runs
Component: Core & setup
Level: 1
Class: Bug fix
Version: 2.2.0i1
In previous versions, the activation housekeeping job was
changed to a background-job running every minute because the
run sometimes took longer than 2 minutes (a hard timeout).
Due to race conditions this change led to errors (missing
activations) in some cases.
Only having it delete activations older than 1h however
led to exhaustion of disk-space in some cases.
This change reintroduces the previously dropped house-keeping
run right before the activation is started, while keeping
the periodically run house-keeping job to clean up afterwards.