non-bloking descriptor after a poll(), don't loop forever on EAGAIN
as poll() may return POLLIN for a descriptor which doesn't have data
to be read. Bump PKGREVISION.
While there add user-destdir support.
run a system with ~100 hosts and 1100+ checks which run every 1-3 mins,
you will have a very terrible experience with this package. See discussion
at:
http://support.nagios.com/forum/viewtopic.php?f=7&t=33274&p=141473#p141473
This patch changes the limit from a whopping three (3) to 128, which is
more reasonable. Though, if others are continuing to have problems with
this package, we could increase it to 1024. Defuzz the rest of the patches.
patches. From: http://nagios.sourceforge.net/docs/nagioscore/4/en/whatsnew.html
Changes and New Features
Performance Improvements:
The performance improvements in Nagios Core 4 come primarily from the following areas:
Core Workers - Core workers are lightweight processes whose only job is to perform checks. Because they are smaller they spawn much more quickly than the the old process which forked the full Nagios Core. In addition, they communicate with the main Nagios Core process using in-memory techniques, eliminating the disk I/O latencies that could previously slow things down, especially in large installations.
Configuration Verification - Configuration verification has been improved so that each configuration item is verified only once. Previously configuration verification was an O(n2) operation.
Event Queue - The event queue now uses a data structure that has O(log n) insertion times versus the O(n) insertion time previously. This means that inserting events into the queue uses much less CPU than in Nagios Core 3.
Macro Resolution - Macros are now sorted on startup so macro lookup can use a binary search. In addition, frequently accessed macros $USERx$, $ARGx$, and $HOSTADDRESS$ are given special case, early lookups.
Object Definitions:
The following changes have been made to object definitions:
The host address attribute is now optional. The address attribute is set to the host name when it is absent. Most configurations set the host name attribute to the DNS host name making the address attribute redundant.
Both hosts and services now support an hourly value attribute. The hourly value attribute is intended to represent the value of a host or service to an organization and is used by the new minimum value contact attribute.
Services now support a parents attribute. A service parent performs a function similar to host parents and can be used in place of service dependencies in simple circumstances.
The failure_prediction_enabled flag has been removed from both host and service object definitions.
Contacts now support a minimum value attribute. The mininum value attribute is used with the host and service hourly value attributes to determine whether to notify a contact on host and service problems.
The host obess_over_host and the service obsess_over_service attributes can now both use the shortened attribute obsess.
Object Behavior:
Contact Inheritance - According to the documentation, contacts should only be inherited from host to service if the service has no other contacts whatsoever (and the same goes for escalations), but the way the code previously worked was that it handled contact_groups and contacts directives separately, meaning services with only 'contacts' specified were still eligible for inheriting 'contact_groups' from the host. This has been updated to comply with the documentation.
Timeperiods - There were several issues processing timeperiods when both exclusions and exceptions were involved. The issues have been corrected.
Configuration:
The following changes have been made to the main Nagios Core configuration, nagios.cfg:
Because there are many ways to obtain object information, the object information is no longer stored if in the object cache if the configuration variable object_cache_file equals '/dev/null'. Setting the variable to '/dev/null' will reduce the disk I/O load.
Because there are many ways to obtain status information, the status information is no longer stored if in the status data file if the configuration variable status_file equals '/dev/null'. Setting the variable to '/dev/null' will reduce the disk I/O load.
There is a new configuration variable, log_current_states, which determines whether current states will be logged in the log files when they are rotated. In Nagios Core 3, this was always the behavior and it is the default in Nagios Core 4. Disabling the logging of current states on log rotation can save considerable disk space for large installations.
There is a new configuration variable, check_workers, which specifies how many worker processes are created when Nagios Core starts. If not specified, the number of worker process is determine by the number of CPUs on the system.
There is a new configuration variable, query_socket, which specifies the location of the query handler socket. The default location is /usr/local/nagios/var/rw/nagios.qh.
The configuration variables, check_result_reaper_frequency and max_check_result_reaper_time, have been deprecated. Because of the new worker architecture, checks are no longer reaped, but they are fed back to core by the worker processes. As a result, these variables no longer make sense.
All file and directory configuration variables in the main nagios.cfg can now use paths that are relative to the location of nagios.cfg.
Although rarely used in the past, creating nagios objects in the main nagios.cfg configuration file was allowed. This is now prohibited.
Macros:
Additions - A new macro, $CHECKSOURCE$, has been added which contains information about what process performed a check.
Changes - If use_large_installation_tweaks is set, the $HOSTGROUPMEMBERS$ and $SERVICEGROUPMEMBERS$ macros are no longer exported because they can consume the available space for environment variables.
Macros are normally available as environment variables when check, event handler, notification, and other commands are run. This can be rather CPU intensive in large Nagios installations, so you can disable the export of environment variables completely with the enable_environment_macros option.
Macro information can be found here.
Query Handler:
The query handler is a general purpose communication mechanism that allows external entities to communicate with Nagios Core in a well-defined manner. As of this writing, all communication with the query handler takes place through a Unix-domain socket whose location is defined by the query_socket configuration variable.
There are currently 5 built-in query handlers.
core - provides Nagios Core management and information
wproc - provides worker process registration, management and information
nerd - provides a subscription service to the Nagios Event Radio Dispatcher (NERD)
help - provides help for the query handler
echo - implements a basic query handler that simply echoes back the queries sent to it
More information about the query handler interface, including an introduction to creating a custom query handler, can be found in the source-supplied documentation.
Core Workers:
Previously, all host and service checks were performed by the full Nagios Core process. This required forking the Nagios Core process for every check. The full Nagios Core process includes a lot of things that are not required to actually perform the check, including check scheduling, downtime handling, processing external commands, etc. As a result, forking the Nagios Core process was much slower than was necessary. When the actual check was run, the forked process again forked a shell to run the check and the shell forked to run the plugin.
In addition, disk files were used as the inter-process communication (IPC) mechanism between the forked Nagios process doing the checking and the main Nagios process handling the check results.
In Nagios Core 4, the process of performing host and service checks is now accomplished using a lightweight worker processes. Standard worker processes start up with the main Nagios Core process and additional, special-purpose workers, can be started at any time after Nagios Core starts. If the check command is "simple" (no shell escapes), the worker process can run the command directly, avoiding the 2 additional forks previously required.
Also in Nagios Core 4, the worker processes report the check results to the main Nagios Core process using in-memory IPC mechanisms (the query handler interface), eliminating the disk I/O bottleneck that used to be an issue in large installations.
When a worker process registers with the main Nagios Core process, it tells Nagios Core what checks it will handle. This feature allows external authors to create special-purpose workers which are optimized to perform certain checks. A sample special-purpose ping check worker is included with the Nagios Core source code in the worker/ping subdirectory.
More information about workers, including an introduction to creating custom workers can be found in the source-supplied documentation.
Nagios Event Radio Dispatcher (NERD):
The Nagios Event Radio Dispatcher (NERD) is a query handler based service that streams Nagios Core events to the subscriber. Currently, there are three channels that can be subscribed to: hostchecks, servicechecks and opathchecks.
libnagios:
libnagios is a library of functions that can be used by developers of query handlers and worker processes. libnagios currently contains the following components.
bitmap - bitmap library for calculating dependency graphs
dkhash - dual-keyed hash api
fanout - sparsely populated array used for downtime, comments, and worker jobs
iobroker - I/O broker library for multiplexing between running tasks and the master nagios process.
iocache - I/O caching libary for bulk-reading requests and parsing them
kvvec - key/value library for parsing requests and building responses
nsock - socket library for connecting to and communicating through the qh socket
nspath - general purpose path library for converting between relative and absolute paths
nsutils - small library with worker related utilities
pqueue - pqueue library written by Volkan Yazici
runcmd - for spawning and reaping commands
skiplist - skiplist library used within Nagios Core
squeue - for maintaining a queue of the running job's timeouts
worker - for utils and stuff nifty to have if you're a worker
Documentation:
Documentation of Nagios Core internals is now provided as part of the source distribution. To create an HTML version of this documentation run 'make dox' from the root of the source distribution tree. The doxygen utilities must be installed to make this documentation.
Tests:
A much more complete test suite is now incuded with the Nagios Core source distribution.
RPM Spec File:
The RPM spec file has been completely overhauled to support more current standards.
Deprecated Features:
Extended Host and Service Information - The hostextinfo and serviceextinfo objects are now deprecated and should not be used. Support for them will be removed in a future version. The same information specified in the hostextinfo and serviceextinfo objects can be specified in the host and service object respectively.
-x/--dont-verify-paths command line option (Don't check for circular object paths) - Because configuration checking is now so much faster, the option to skip checking for circular object paths has been deprecated.
The following configuration variables have been deprecated: check_result_reaper_frequency, max_check_result_reaper_time, sleep_time, external_command_buffer_slots, command_check_interval
Obsoleted Features:
Failure Prediction - As noted above, the failure_prediction_enabled flag has been removed from both host and service object definitions. Failure predition was never fully implemented and would require breaking the paradigm that Nagios Core knows nothing about the performance data returned by plugins. Failure prediction is much more approprately handled by an add-on than by Nagios Core.
-o/--dont-verify-objects command line option - This option, while accepted in Nagios Core 3, has neither been advertized nor has had any effect for quite some time. The option has been removed in Nagios Core 4.
Embedded Perl - Embedded Perl has historically been the least tested and the most problem prone part of Nagios Core. A significant part of the issue is that there are so many versions of Perl available. The performance enhancements provided by the new worker process architecture make up for any performance loss due to the removal of embeddd Perl. In addition, the worker process architecture makes possible the implementation of a special purpose worker to persistently load and run Perl plugins. The following configuration variables that were related to embedded Perl have been obsoleted: use_embedded_perl_implicitly, enable_embedded_perl, p1_file.
Miscellaneous:
Object IDs - Primarily only of interest to developers, all of the first-class objects now have object IDs. First-class objects are timeperiod, command, contact, host, service, escalations, dependencies and all kinds of groups. Object IDs are not persistent and are recreated on each restart.
Do it for all packages that
* mention perl, or
* have a directory name starting with p5-*, or
* depend on a package starting with p5-
like last time, for 5.18, where this didn't lead to complaints.
Let me know if you have any this time.
Upstream changes:
* Added handler for SIGXFSZ signal (Eric Stanley)
* Fixed bug #444: Nagios 3.5.0 problem with macro $ADMINEMAIL$ :
@ is converted to %40 after 2 nagios reload (Duplicate of bug #407)
* Fixed bug #407: Reloading nagios config causes spaces in
notifications to become plus signs (Alexey Dvoryanchikov)
* Fixed bug #445: Adding triggered downtime for child hosts causes
a SIGSEGV on restart/reload (Eric Stanley)
* Fixed bug #375: Freshness expiration never reached and bug #427:
freshness threshold doesn't work if it is set long (Scott Wilkerson,
Eric Stanley)
* Fixed bug #432: Downtime scheduled as "Nagios Process" and not
the Users name (Sam Lansing, Eric Stanley)
off-by-one errors causing information leakage and possibly DoS.
Restructure the patch files to follow the newer naming conventions.
Add the rc.d script to PLIST.
Bump PKGREVISION.
based on PR pkg/48082.
3.5.0 - 03/15/2013
------------------
* Fixed bug #403: The "configuration" page of the webui doesn't use entity
encoding when displaying the "command expansion" item (Eric Stanley, Phil
Randal)
* Fixed bug #424: Nagios Core 3.4.4 seg fault (core dump) on restart after
removing config for running service (Eric Stanley)
* Updated CGI utility functions to support UTF-8 characters (Eric Stanley)
* Fixed bug where selecting Command Expansion from Configuration CGI page would
display commands instead (Eric Stanley)
* Fixed bug #369: status.cgi crashes with segfault when there are german ulauts
(äöüß) in the hostname or the servicename (Eric Stanley)
* Fixed bug #418: Scheduled Downtime Notifications Resent On Nagios
Restart/reload (Eric Stanley)
to address issues with NetBSD-6(and earlier)'s fontconfig not being
new enough for pango.
While doing that, also bump freetype2 dependency to current pkgsrc
version.
Suggested by tron in PR 47882
a) refer 'perl' in their Makefile, or
b) have a directory name of p5-*, or
c) have any dependency on any p5-* package
Like last time, where this caused no complaints.
Changelog since 3.3.1
3.4.3 - 11/30/2012
------------------
- Reverted squeue changes intended for Nagios 4
- Reapplied all patches from 3.4.2 release
- Applied fix for pagination and sorting on status.cgi #381 (Phil Randal)
3.4.2 - 11/09/2012
------------------
FIXES
* Fixed issue where deleting a downtime could cause Nagios to crash (Eric Stanley)
* Corrected logic so that end times for flexible downtimes are calculated from the dow
ntime start rather than the current time in the case where Nagios is restarted (Eric S
tanley)
* Fixed issue introduced by fix for bug #124 where flexible downtimes are not taken in
to account on Nagios restart. (Scott Wilkerson, Eric Stanley)
* Fixed bug #247: If a service reports no performance data, the perfdata log file has
no line indicating the test. (omnikron@free.fr)
* Fixed link for unhandled unreachable host problems on tactical overview page (Rudolf
Cejka)
* Fixed bug #345 with wild card searches not paging properly on status.cgi (Phil Randa
l)
* Fixed bug #343 on status.cgi where Service Group Summary can potentially show wrong
totals (Mark Ziesemer)
* Fixed memory leaks on SIGHUP (Carlos Velasco)
3.4.1 - 05/11/2012
------------------
FIXES
* Double quotes in check_command definition break functionality (#332, reverts #86)
3.4.0 - 05/04/2012
------------------
ENHANCEMENTS
* Added service_check_timeout_state configuration variable (Bill McGonigle)
* Permanently remove sleep on run_event == FALSE in main loop (Max <perldork@webwizard
design.com>)
* Reduce notification load by moving notification viability check into notification li
st creation (Opsview Team)
* Added code to apply allow_empty_hostgroup_assignment flag to host and service depend
encies (Daniel Wittenberg)
* Users can now see hostgroups and servicegroups that contain at least one host or ser
vice they are authorized for, instead of having to be authorized for them all (Ethan G
alstad)
* RSS feed boxes fallback if an error occurs (Ethan Galstad)
* RSS feeds no longer block main page load (Mike Guthrie)
FIXES
* Fix $NOTIFICATIONRECIPIENTS$ macro to contain all contacts assigned to host|service,
not only notified contacts (Bug #98 Matt Harrington)
* Scheduled Downtime Notifications Resent On Nagios Restart/reload (Bug #124 - ricardo
)
* NOTIFICATIONTYPE MACRO never became CUSTOM (Bug #168 - Alexey Dvoryanchikov)
* Plugged minor memory leaks in notification logic
Changes since 3.2.3:
ENHANCEMENTS
* Added support for same host service dependencies with servicegroups (Mathieu Gagné)
* Empty hostgroups referenced from services now optionally generate a warning instead of an error.
* Documentation links now point to online resources
* Matt Wall's Exfoliation theme is now installed by default. You can reinstall the classic theme with "make install-classicui"
* Downtime delete commands made "distributable" by deleting by host group name, host name or start time/comment (Opsview team)
* Allow status.cgi to order by "host urgency" (Jochen Bern)
* Added news items and quick links to main splash page
* Added ability to authenticate to CGIs using contactgroup name (Stephen Gran)
FIXES
* Fixes status.cgi when called with no parameters, where host should be set to all if none specified (Michael Friedrich)
* Fixes possible validation error with empty hostgroups/servicegroups (Sven-Göran Bergh)
* Performance-data handling and checking is now thread-safe so long as embedded perl is not used.
* Children should no longer hang on mutex locks held in parent for localtime() (and similar) calls.
* Debug logging is now properly serialized, using soft-locking with a timeout of 150 milliseconds to avoid multiple threads competing for the privilege to write debug info.
* Fixed extraneous alerts for services when host is down
* Fixed incorrect parsing of multi-line host check results (Jochen Bern)
* Fixed bug with passive host checks being incorrectly sent to event brokers as active checks
* Fixed bug where passive host check status updates were not being propagated to event brokers
* Reverted 'Fix for retaining host display name and alias, as well as service display name' as configuration information stored incorrectly over a reload
* Fixed compile warnings for size_t (Michael Friedrich)
* Fixed problem where acknowledgements were getting reset when a hard state change occurred
* Removed duplicated unlinks for check result files with multiple results
* Fixed race condition on flexible downtime commands when duration not set or zero (Michael Friedrich)
* Fixed flexible downtime on service hard state change doesn't get triggered/activated (Michael Friedrich)
* Fixed XSS vulnerability in config.cgi and statusmap.cgi (Stefan Schurtz)
* Fixed segfault when sending host notifications (Michael Friedrich)
* Fixed bug where unauthorized contacts could issue hostgroup and servicegroup commands (Sven Nierlein)
ChangeLog:
* Fixes problem where disabling all active hosts/services was not taking effect
* Fixes for compiler warnings (code cleanup by Stephen Gran)
* Fixes for format errors in event handler logging (Guillaume Rousse)
* Fixed incorrect info in sample nagios.cfg file for state_retention_file (Michael Friedrich)
* Fixed broker_event_handler() to return ERR if data is NULL (Michael Friedrich)
* Patch to new_mini_epn to allow any command line length without breaking on extra trailing or leading whitespace (Ray Bengen)
* Patch to mini_epn to allow any command line length (Thomas Guyot-Sionnest)
* Patch to speed up loading of state retention data (Matthieu Kermagoret)
* Custom notifications are now suppressed during scheduled downtime (Sven Nierlein)
* Added code to warn user about exit code of 126 meaning plugin is not executable (bug #153)
* Scheduled downtime can now start on SOFT error states (bug #47)
* Main window frame URL can now be specify with a "corewindow=" parameter
* Improved config CGI shows commands, command args in an easier to use manner (Jochen Bern)
* Added ability for NEB modules to override execution of event handlers (Sven Nierlein)
* Custom macros are no longer cleaned/stripped as they are user-defined and should be trusted (Peter Morch)
* Fix for choosing next valid time on day of DST change when clocks go one hour backwards
* Fix for nagios now erroring when "Error: Could not find any contactgroup matching..." displayed
* Fix tap tests for Sol0 and newer versions of Test::Harness
* Fix for notifications not being sent out when scheduled downtime is canceluzzner)
* Fix for first notification delay being calculated incorrectly, and notifications potentially going out early (Plachowski)
* Fix for text of scheduling downtime of all services on a host (Holger Weiss)
* Fix for services inheriting notification period from hosts if not defined (Gordon Messmer)
* Fix for incorrect service states on host failures (bug #130 Pet)
* Fix for incorrect service state attributes being set on host failures (bug #128 Petya Kohts)
* Fix for non-scheduled hostsnd services not being updated in NDOUtils
* Fix for typos in TAC, CMD CGIs (bugs #150, #144, #148)
* Fix for types in documentation (bugs #145, #105, #106)
* Fix for incorrect host state counts in status CGI when viewing servicegroups (bug #72)
* Fix few Splunk integration query parameters (bug #136)
* Fix for extra field header in availability CSV export (bug #113)
* Fix foracro processing code modifying input string (Jochen Bern)
* Fix for update check API
* Fix for CGI speedup when persistent=0 f comments
* Fix for event execution loop re-scheduling host checks instead of executing them if service checks are disabled (b #152)
* Fix for segfaults on Solaris (Torsten Huebler)
* Fix for incorrect comment expiration times being passed to event bror (Mattieu Kermagot)
* Doc updates related to cleaning of custom macros (Peter Valdemar Morch)
* Fix to sample notify-service--email command (bug #62)
* Fix for retaining host display name and alias, as well as service display name (Folkert van Heusden* Link to allow scheduling downtime for all services on a host (Hendrik Baecker)
* Speedup to CGIs when lots of comments or dotimes in status.dat file (Jonathan Kamens)
* Patch for new_mini_epn to allow for any command line length without breaking extra trailing or leading whitespace (Ray Bengen)
* Fix for incorrect scheduling when time has gone back an hour (partial fix for 24x7)
* Fix for compile on Fedora Core 3 (bug #0000082)
* Fix for compile on Solaris
* Fix for logging test, which was not timezone aware (bug #0000077 - Allan Clark)
* Trivial cleanups for autoconf (Allan Clark)
* Fix for CSS validation of padding: X
* Fix for documentation re: case-insensitive nature of custom variables (Marc Powell)
* Fix for template configurations which use negated wildcards (Tim Wilde)
* Fix for read-only permissions bug in CGIs that caused problems viewing comments (bug #0000029)
* Fix for incorrect CGI reports (availability, trends, etc.) when reporting period spans Daylight Savings Time (bug #0000046)
* Fix for detection of truecolor support in GD library (Lars Hecking)
* Reverted to use --datadir configure script option instead of the more recently introduced --datarootdir option
* Status and retention files are now flushed/synced to disk to prevent incomplete information being displayed in CGIs
* Fix for incorrect next service check time calculation when Nagios is reloaded with different timeperiod ranges
* Updated Fedora quistart guide to indicate PHP requirements
* Known issue: Service checks that are defined with timeperiods that contain "exclude" directives are incorrectly re-scheduled. Don't use these for now - we'll get this
fixed for 3.4
to trigger/signal a rebuild for the transition 5.10.1 -> 5.12.1.
The list of packages is computed by finding all packages which end
up having either of PERL5_USE_PACKLIST, BUILDLINK_API_DEPENDS.perl,
or PERL5_PACKLIST defined in their make setup (tested via
"make show-vars VARNAMES=..."), minus the packages updated after
the perl package update.
sno@ was right after all, obache@ kindly asked and he@ led the
way. Thanks!
