ID: 8124
Title: Avoid configuration changes to be become active without activate changes
Component: config
Level: 2
Class: Bug fix
Version: 1.2.7i1
The Check_MK keepalive mode, which is used for the Check_MK helpers and for the
RRD creation helper, now uses a new method for reading the configuration from
<tt>main.mk</tt>, the other <tt>*.mk</tt> files and the <tt>autochecks</tt>. This
new method has two advantages:
<ul>
<li>Changes done to <tt>*.mk</tt> files can no longer become active unless they
are activated. This even holds for core or server restarts and for crashed check
helpers</li>
<li>Check helpers and RRD helper do not any longer need to restart themselves
after a configuration change but simply reload the changed configuration. This
has a muss smaller CPU footprint when activating changes.</li>
<li>Autochecks files are now only read for the host that is currently being
delt with. This makes the internal check table much smaller and speeds
configuration reloads.</li>
</ul>
The gross result is that configuration changes are now really automic and
also faster.
H2:Implementation details
Whenever you create the configuration for the core (options <tt>-B</tt>, <tt>-O</tt>,
<tt>-R</tt> or <tt>-U</tt> or <i>Activate Changes</i> via WATO) then the complete
configuration (the result from parsing <tt>main.mk</tt> and friends) is written to
the file <tt>var/check_mk/core/config.mk</tt>. The core then triggers the check
helpers for configuration reload. The helpers simply re-read that file and are
immediately up-to-date. At the same time copies (hard links) of the autochecks
files in <tt>var/check_mk/autochecks</tt> are being created in <tt>var/check_mk/core/autochecks</tt>.
These files are used during the monitoring.
Note: The <i>Discovery</i> check (formerly known as <i>Inventory<i> check) does <i>not</i>
use the activated configuration but still that one that is modified via WATO. That way
after doing a service discovery the discovery check will immediately be happy - without
a core restart. This is just as it used to be.
ID: 8118
Title: Smartping: Fixed "no buffer space available" error when having a lot of hosts configured
Component: cmc
Level: 2
Class: Bug fix
Version: 1.2.7i1
We already changed the way smartping was operating some time ago, this solved the
"no buffer space available" problem for many systems and increased the possible
number of hosts to be pinged dramatically.
This did not fix the problem for all systems. Some systems with maybe slower network
cards or drivers still had the problem that the message "no buffer space available"
appeared from time to time in the cmc.log. This occured when the cmc was asking for
ping requests faster than the kernel / NIC could process them. The send queue of the
single raw socket was filling up to the maximum and then resulting in this error message.
The icmpsender now opens 8 sockets instead of 1 for sending the ICMP requests.
Additionally the socktes now get a larger send buffer to make it less likely to fill up.
ID: 8113
Title: Timeouts are now handled like classic SNMP implementation
Component: inline-snmp
Level: 2
Class: Bug fix
Version: 1.2.6b1
Timeouts of walks/bulkwalks of Inline-SNMP are now handled equal
to classic SNMP implementation. When a timeout occurs for the
configured X SNMP retries after Y SNMP timeouts (configured via
SNMP Timing Options), the whole "Check_MK" service is terminated
assuming that the host is not available.
ID: 8117
Title: The "Check_MK Discovery" service is executed in a much more performant way
Component: cmc
Level: 2
Class: Bug fix
Version: 1.2.7i1
The "Check_MK Discovery" check was executed as active check, initializing the whole
python interpreter, check_mk and the whole check_mk configuration for each check,
just to perform an inventory check for a single host.
Now the "Check_MK Discovery" service also uses the permanently running Check_MK
Helper processes just like the regular Check_MK service which results in a big
performanc boosts for environments with a large number of hosts using the
"Check_MK Discovery" service.
ID: 8111
Title: Fix exception in agent bakery if host has been removed
Component: agents
Level: 2
Class: Bug fix
Version: 1.2.6b1
When a host has been removed or reclassified as having no Check_MK
agent then the page showing all agent configuration raises
and exception. This has been fixed.
ID: 8112
Title: Fix exception due to reporting module for RedHat 5.X
Component: Reporting & Availability
Level: 2
Class: Bug fix
Version: 1.2.7i1
Problem was a Python syntax of <tt>class Document():</tt> that is
not allowed in Python 2.4.
ID: 8109
Title: Report scheduler: fix instant emailing of report in case period is not daily
Component: Reporting & Availability
Level: 2
Class: Bug fix
Version: 1.2.6b1
ID: 8105
Title: Fix exception when adding a PNP graph for a single service to a report
Component: Reporting & Availability
Level: 2
Class: Bug fix
Version: 1.2.6b1
ID: 8106
Title: Fix exceptions in reports that contain Event Console views
Component: Reporting & Availability
Level: 2
Class: Bug fix
Version: 1.2.6b1
The problem were columns that contained numbers that were not converted
to strings correctly.