icinga2/doc/16-upgrading-icinga-2.md

44 KiB

Upgrading Icinga 2

Upgrading Icinga 2 is usually quite straightforward. Ordinarily the only manual steps involved are scheme updates for the IDO database.

Specific version upgrades are described below. Please note that version updates are incremental. An upgrade from v2.6 to v2.8 requires to follow the instructions for v2.7 too.

Upgrading to v2.14

Dependencies and Redundancy Groups

Before Icinga v2.12 all dependencies were cumulative. I.e. the child was considered reachable only if no dependency was violated. In v2.12 and v2.13, all dependencies were redundant. I.e. the child was considered unreachable only if no dependency was fulfilled.

v2.14 restores the pre-v2.12 behavior, but allows to override it. I.e. you can still make any number of your dependencies redundant, as you wish. For details read the docs' redundancy groups section.

Email Notification Scripts

The email notification scripts shipped with Icinga 2 (/etc/icinga2/scripts) now link to Icinga DB Web, not the monitoring module. Both new and existing installations are affected unless you've altered the scripts.

In the latter case package managers won't upgrade those "config" files in-place, but just put files with similar names into the same directory. This allows you to patch them by yourself based on diff(1).

On the other hand, if you want to stick to the monitoring module for now, add any comments to the notification scripts before upgrading. This way package managers won't touch those files.

Upgrading to v2.13

DB IDO Schema Update

There is an optional schema update on MySQL which increases the max length of object names from 128 to 255 characters.

Please proceed here for the MySQL upgrading docs.

Behavior changes

Deletion of child downtimes on services

Service downtimes created while using the all_services flag on the schedule-downtime API action will now automatically be deleted when deleting the hosts downtime.

Windows Event Log

Icinga 2.13 now supports logging to the Windows Event Log. Icinga will now also log messages from the early startup phase to the Windows Event Log. These were previously missing from the log file and you could only see them by manually starting Icinga in the foreground.

This feature is now enabled and replaces the existing mainlog feature logging to a file. When upgrading, the installer will enable the windowseventlog feature and disable the mainlog feature. Logging to a file is still possible. If you don't want this configuration migration on upgrade, you can opt-out by installing the %ProgramData%\icinga2\etc\icinga2\features-available\windowseventlog.conf file before upgrading to Icinga 2.13.

Broken API package name validation

This version has replaced a broken regex in the API package validation code which results in package names now being validated correctly. Package names should now only consist of alphanumeric characters, dashes and underscores.

This change only applies to newly created packages to support already existing ones.

Upgrading to v2.12

Behavior changes

The behavior of multi parent dependencies was fixed to e.g. render hosts unreachable when both router uplinks are down.

Previous behaviour:

  1. parentHost1 DOWN, parentHost2 UP => childHost not reachable
  2. parentHost1 DOWN, parentHost2 DOWN => childHost not reachable

New behavior:

  1. parentHost1 DOWN, parentHost2 UP => childHost reachable
  2. parentHost1 DOWN, parentHost2 DOWN => childHost not reachable

Please review your Dependency configuration as 1) may lead to different results for

  • last_reachable via REST API query
  • Notifications not suppressed by faulty reachability calculation anymore

Breaking changes

As of v2.12 our API URL endpoint /v1/actions/acknowledge-problem refuses acknowledging an already acknowledged checkable by overwriting the acknowledgement. To replace an acknowledgement you have to remove the old one before adding the new one.

The deprecated parameters --cert and --key for the pki save-cert CLI command have been removed from the command and documentation.

Upgrading to v2.11

Bugfixes for 2.11

2.11.1 on agents/satellites fixes a problem where 2.10.x as config master would send out an unwanted config marker file, thus rendering the agent to think it is autoritative for the config, and never accepting any new config files for the zone(s). If your config master is 2.11.x already, you are not affected by this problem.

In order to fix this, upgrade to at least 2.11.1, and purge away the local config sync storage once, then restart.

yum install icinga2

rm -rf /var/lib/icinga2/api/zones/*
rm -rf /var/lib/icinga2/api/zones-stage/*

systemctl restart icinga2

2.11.2 fixes a problem where the newly introduced config sync "check-change-then-reload" functionality could cause endless reload loops with agents. The most visible parts are failing command endpoint checks with "not connected" UNKNOWN state. Only applies to HA enabled zones with 2 masters and/or 2 satellites.

In order to fix this, upgrade all agents/satellites to at least 2.11.2 and restart them.

Packages

EOL distributions where no packages are available with this release:

  • SLES 11
  • Ubuntu 14 LTS
  • RHEL/CentOS 6 x86

Raspbian Packages are available inside the icinga-buster repository on https://packages.icinga.com. Please note that Stretch is not supported suffering from compiler regressions. Upgrade to Raspbian Buster is highly recommended.

Added: Boost 1.66+

The rewrite of our core network stack for cluster and REST API requires newer Boost versions, specifically >= 1.66. For technical details, please continue reading in this issue.

Distribution Repository providing Boost Dependencies
CentOS 7 EPEL repository
RHEL 7 EPEL repository
RHEL/CentOS 6 x64 packages.icinga.com
Fedora Fedora Upstream
Debian 10 Buster Debian Upstream
Debian 9 Stretch Backports repository New since 2.11
Debian 8 Jessie packages.icinga.com
Ubuntu 18 Bionic packages.icinga.com
Ubuntu 16 Xenial packages.icinga.com
SLES 15 SUSE Upstream
SLES 12 packages.icinga.com (replaces the SDK repository requirement)

On platforms where EPEL or Backports cannot satisfy this dependency, we provide Boost as package on our package repository for your convenience.

After upgrade, you may remove the old Boost packages (1.53 or anything above) if you don't need them anymore.

Added: .NET Framework 4.6

We modernized the graphical Windows wizard to use the more recent .NET Framework 4.6. This requires that Windows versions older than Windows 10/Windows Server 2016 installs at least .NET Framework 4.6. Starting with Windows 10/Windows Server 2016 a .NET Framework 4.6 or higher is installed by default.

The MSI-Installer package checks if the .NET Framework 4.6 or higher is present, if not the installation wizard will abort with an error message telling you to install at least .NET Framework 4.6.

Removed: YAJL

Our JSON library, namely YAJL, isn't maintained anymore and may cause crashes.

It is replaced by JSON for Modern C++ by Niels Lohmann and compiled into the binary as header only include. It helps our way to C++11 and allows to fix additional UTF8 issues more easily. Read more about its design goals and benchmarks.

Core

Reload Handling

2.11 provides fixes for unwanted notifications during restarts. The updated systemd service file now uses the KillMode=mixed setting.

The reload handling was improved with an umbrella process, which means that normal runtime operations include 3 processes. You may need to adjust the local instance monitoring of the procs check.

More details can be found in the technical concepts chapter.

Downtime Notifications

Imagine that a host/service changes to a HARD NOT-OK state, and its check interval is set to a high interval e.g. 1 hour.

A maintenance downtime prevents the notification being sent, but once it ends and the host/service is still in a downtime, no immediate notification is re-sent but you'll have to wait for the next check.

Another scenario is with one-shot notifications (interval=0) which would never notify again after the downtime ends and the problem state being intact. The state change logic requires to recover and become HARD NOT-OK to notify again.

In order to solve these problems with filtered/suppressed notifications in downtimes, v2.11 changes the behaviour like this:

  • If there was a notification suppressed in a downtime, the core stores that information
  • Once the downtime ends and the problem state is still intact, Icinga checks whether a re-notification should be sent immediately

A new cluster message was added to keep this in sync amongst HA masters.

Important

In order to properly use this new feature, all involved endpoints must be upgraded to v2.11.

Network Stack

The core network stack has been rewritten in 2.11 (some say this could be Icinga 3).

You can read the full story here.

The only visible changes for users are:

  • No more dead-locks with hanging TLS connections (Cluster, REST API)
  • Better log messages in error cases
  • More robust and stable with using external libraries instead of self-written socket I/O

Coming with this release, we've also updated TLS specific requirements explained below.

TLS 1.2

v2.11 raises the minimum required TLS version to 1.2. This is available since OpenSSL 1.0.1 (EL6 & Debian Jessie).

Older Icinga satellites/agents need to support TLS 1.2 during the TLS handshake.

The api feature attribute tls_protocolmin now only supports the value TLSv1.2 being the default.

Hardened Cipher List

The previous default cipher list allowed weak ciphers. There's no sane way other than explicitly setting the allowed ciphers.

The new default sets this to:

ECDHE-ECDSA-AES256-GCM-SHA384:ECDHE-RSA-AES256-GCM-SHA384:ECDHE-ECDSA-CHACHA20-POLY1305:ECDHE-RSA-CHACHA20-POLY1305:ECDHE-ECDSA-AES128-GCM-SHA256:ECDHE-RSA-AES128-GCM-SHA256:ECDHE-ECDSA-AES256-SHA384:ECDHE-RSA-AES256-SHA384:ECDHE-ECDSA-AES128-SHA256:ECDHE-RSA-AES128-SHA256:DHE-RSA-AES128-GCM-SHA256:DHE-RSA-AES256-GCM-SHA384:AES256-GCM-SHA384:AES128-GCM-SHA256

You can override this setting in the api feature with the cipher_list attribute.

In case that one of these ciphers is marked as insecure in the future, please let us know with an issue on GitHub.

Cluster

Agent Hosts with Command Endpoint require a Zone

2.11 fixes bugs where agent host checks would never be scheduled on the master. One definite requirement is that the checkable host/service is put into a zone.

By default, the Director puts the agent host in zones.d/master and you're good to go. If you manually manage the configuration, the config compiler now throws an error with command_endpoint being set but no zone defined.

The most convenient way with e.g. managing the objects in conf.d is to move them into the master zone. Please continue in the troubleshooting docs for further instructions.

Config Sync

2.11 overhauls the cluster config sync in many ways. This includes the following under the hood:

  • Synced configuration files are not immediately put into production, but left inside a stage.
  • Unsuccessful config validation never puts the config into production, additional logging and API states are available.
  • Zone directories which are not configured in zones.conf, are not included anymore on secondary master/satellites/clients.
  • Synced config change calculation use checksums instead of timestamps to trigger validation/reload. This is more safe, and the usage of timestamps is now deprecated.
  • Don't allow parallel cluster syncs to avoid race conditions with overridden files.
  • Deleted directories and files are now purged, previous versions had a bug.

Whenever a newer child endpoint receives a configuration update without checksums, it will log a warning.

Received configuration update without checksums from parent endpoint satellite1. This behaviour is deprecated. Please upgrade the parent endpoint to 2.11+

This is a gentle reminder to upgrade the master and satellites first, prior to installing new clients/agents.

Technical details are available in the technical concepts chapter.

Since the config sync change detection now uses checksums, this may fail with anything else than syncing configuration text files. Syncing binary files were never supported, but rumors say that some users do so.

This is now prohibited and logged.

[2019-08-02 16:03:19 +0200] critical/ApiListener: Ignoring file '/etc/icinga2/zones.d/global-templates/forbidden.exe' for cluster config sync: Does not contain valid UTF8. Binary files are not supported.
Context:
	(0) Creating config update for file '/etc/icinga2/zones.d/global-templates/forbidden.exe'
	(1) Activating object 'api' of type 'ApiListener'

Such binaries wrapped into JSON-RPC cluster messages may always cause changes and trigger reload loops. In order to prevent such harm in production, use infrastructure tools such as Foreman, Puppet, Ansible, etc. to install plugins on the masters, satellites and agents.

Config Sync: Zones in Zones

The cluster config sync works in the way that configuration put into /etc/icinga2/zones.d only is included when configured outside in /etc/icinga2/zones.conf.

If you for example create a "Zone Inception" with defining the satellite zone in zones.d/master, the config compiler does not re-run and include this zone config recursively from zones.d/satellite.

Since v2.11, the config compiler is only including directories where a zone has been configured. Otherwise it would include renamed old zones, broken zones, etc. and those long-lasting bugs have been now fixed.

Please consult the troubleshoot docs for concrete examples and solutions.

HA-aware Features

v2.11 introduces additional HA functionality similar to the DB IDO feature. This enables the feature being active only on one endpoint while the other endpoint is paused. When one endpoint is shut down, automatic failover happens.

This feature is turned off by default keeping the current behaviour. If you need it active on just one endpoint, set enable_ha = true on both endpoints in the feature configuration.

This affects the following features:

HA Failover

The reconnect failover has been improved, and the default failover_timeout for the DB IDO features has been lowered from 60 to 30 seconds. Object authority updates (required for balancing in the cluster) happen more frequenty (was 30, is 10 seconds). Also the cold startup without object authority updates has been reduced from 60 to 30 seconds. This is to allow cluster reconnects (lowered from 60s to 10s in 2.10) before actually considering a failover/split brain scenario.

The IdoMysqlConnection and IdoPgsqlConnection objects provide a new attribute named last_failover which shows the last failover timestamp. This value also is available in the ido CheckCommand output.

CLI Commands

The troubleshoot CLI command has been removed. It was never completed, and turned out not to provide required details for GitHub issues anyways.

We didn't ask nor endorse users on GitHub/Discourse in the past 2 years, so we're removing it without deprecation.

Issue templates, the troubleshooting docs and support knowledge has proven to be better.

Permissions

CLI commands such as api setup, node wizard/setup, feature enable/disable/list required root permissions previously. Since the file permissions allow the Icinga user to change things already, and users kept asking to run Icinga on their own webspace without root permissions, this is now possible with 2.11.

If you are running the commands with a different user than the compiled ICINGA_USER and ICINGA_GROUP CMake settings (icinga everywhere, except Debian with nagios for historical reasons), ensure that this user has the capabilities to change to a different user.

If you still encounter problems, run the aforementioned CLI commands as root, or with sudo.

CA List Behaviour Change

ca list only shows the pending certificate signing requests by default.

You can use the new --all parameter to show all signing requests. Note that Icinga automatically purges signed requests older than 1 week.

New: CA Remove/Restore

ca remove allows you to remove pending signing requests. Once the master receives a CSR, and it is marked as removed, the request is denied.

ca restore allows you to restore a removed signing request. You can list removed signing requests with the new --removed parameter for ca list.

Configuration

The deprecated concurrent_checks attribute in the checker feature has no effect anymore if set. Please use the MaxConcurrentChecks constant in constants.conf instead.

REST API

Actions

The schedule-downtime action supports the all_services parameter for Host types. Defaults to false.

Config Packages

Deployed configuration packages require an active stage, with many previous allowed. This mechanism is used by the Icinga Director as external consumer, and Icinga itself for storing runtime created objects inside the _api package.

This includes downtimes and comments, which where sometimes stored in the wrong directory path, because the active-stage file was empty/truncated/unreadable at this point.

2.11 makes this mechanism more stable and detects broken config packages. It will also attempt to fix them, the following log entry is perfectly fine.

[2019-05-10 12:12:09 +0200] information/ConfigObjectUtility: Repairing config package '_api' with stage 'dbe0bef8-c72c-4cc9-9779-da7c4527c5b2'.

If you still encounter problems, please follow this troubleshooting entry.

DB IDO MySQL Schema

The schema for MySQL contains an optional update which drops unneeded indexes. You don't necessarily need to apply this update.

Documentation

  • Custom attributes have been renamed to Custom variables following the name vars and their usage in backends and web interfaces. The term custom attribute still applies, but referring from the web to the core docs is easier.
  • The distributed environment term client has been refined into agent. Wordings and images have been adjusted, and a client only is used as general term when requesting something from a parent server role.
  • The images for basics, modes and scenarios in the distributed monitoring chapter have been re-created from scratch.
  • 02-getting-started.md was renamed to 02-installation.md, 04-configuring-icinga-2.md into 04-configuration.md. Apache redirects will be in place.

Upgrading to v2.10

Path Constant Changes

During package upgrades you may see a notice that the configuration content of features has changed. This is due to a more general approach with path constants in v2.10.

The known constants SysconfDir and LocalStateDir stay intact and won't break on upgrade. If you are using these constants in your own custom command definitions or other objects, you are advised to revise them and update them according to the documentation.

Example diff:

object NotificationCommand "mail-service-notification" {
-  command = [ SysconfDir + "/icinga2/scripts/mail-service-notification.sh" ]
+  command = [ ConfigDir + "/scripts/mail-service-notification.sh" ]

If you have the ICINGA2_RUN_DIR environment variable configured in the sysconfig file, you need to rename it to ICINGA2_INIT_RUN_DIR. ICINGA2_STATE_DIR has been removed and this setting has no effect.

Note

This is important if you rely on the sysconfig configuration in your own scripts.

New Constants

New Icinga constants have been added in this release.

  • Environment for specifying the Icinga environment. Defaults to not set.
  • ApiBindHost and ApiBindPort to allow overriding the default ApiListener values. This will be used for an Icinga addon only.

Configuration: Namespaces

The keywords namespace and using are now reserved for the namespace functionality provided with v2.10. Read more about how it works here.

Configuration: ApiListener

Anonymous JSON-RPC connections in the cluster can now be configured with max_anonymous_clients attribute. The corresponding REST API results from /v1/status/ApiListener in json_rpc have been renamed from clients to anonymous_clients to better reflect their purpose. Authenticated clients are counted as connected endpoints. A similar change is there for the performance data metrics.

The TLS handshake timeout defaults to 10 seconds since v2.8.2. This can now be configured with the configuration attribute tls_handshake_timeout. Beware of performance issues with setting a too high value.

API: schedule-downtime Action

The attribute child_options was previously accepting 0,1,2 for specific child downtime settings. This behaviour stays intact, but the new proposed way are specific constants as values (DowntimeNoChildren, DowntimeTriggeredChildren, DowntimeNonTriggeredChildren).

Notifications: Recovery and Acknowledgement

When a user should be notified on Problem and Acknowledgement, v2.10 now checks during the Acknowledgement notification event whether this user has been notified about a problem before.

  types = [ Problem, Acknowledgement, Recovery ]

If no Problem notification was sent, and the types filter includes problems for this user, the Acknowledgement notification is not sent.

In contrast to that, the following configuration always sends Acknowledgement notifications.

  types = [ Acknowledgement, Recovery ]

This change also restores the old behaviour for Recovery notifications. The above configuration leaving out the Problem type can be used to only receive recovery notifications. If Problem is added to the types again, Icinga 2 checks whether it has notified a user of a problem when sending the recovery notification.

More details can be found in this PR.

Stricter configuration validation

Some config errors are now fatal. While it never worked before, icinga2 refuses to start now!

For example the following started to give a fatal error in 2.10:

  object Zone "XXX" {
    endpoints = [ "master-server" ]
    parent = "global-templates"
  }

critical/config: Error: Zone 'XXX' can not have a global zone as parent.

Package Changes

Debian/Ubuntu drops the libicinga2 package. apt-get upgrade icinga2 won't remove such packages leaving the upgrade in an unsatisfied state.

Please use apt-get full-upgrade or apt-get dist-upgrade instead, as explained here.

On RHEL/CentOS/Fedora, icinga2-libs has been obsoleted. Unfortunately yum's dependency resolver doesn't allow to install older versions than 2.10 then. Please read here for details.

RPM packages dropped the Classic UI package in v2.8, Debian/Ubuntu packages were forgotten. This is now the case with this release. Icinga 1.x is EOL by the end of 2018, plan your migration to Icinga Web 2.

Upgrading to v2.9

Deprecation and Removal Notes

  • Deprecation of 1.x compatibility features: StatusDataWriter, CompatLogger, CheckResultReader. Their removal is scheduled for 2.11. Icinga 1.x is EOL and will be out of support by the end of 2018.
  • Removal of Icinga Studio. It always has been experimental and did not satisfy our high quality standards. We've therefore removed it.

Sysconfig Changes

The security fixes in v2.8.2 required moving specific runtime settings into the Sysconfig file and environment. This included that Icinga 2 would itself parse this file and read the required variables. This has generated numerous false-positive log messages and led to many support questions. v2.9.0 changes this in the standard way to read these variables from the environment, and use sane compile-time defaults.

Note

In order to upgrade, remove everything in the sysconfig file and re-apply your changes.

There is a bug with existing sysconfig files where path variables are not expanded because systemd does not support shell variable expansion. This worked with SysVInit though.

Edit the sysconfig file and either remove everything, or edit this line on RHEL 7. Modify the path for other distributions.

vim /etc/sysconfig/icinga2

-ICINGA2_PID_FILE=$ICINGA2_RUN_DIR/icinga2/icinga2.pid
+ICINGA2_PID_FILE=/run/icinga2/icinga2.pid

If you want to adjust the number of open files for the Icinga application for example, you would just add this setting like this on RHEL 7:

vim /etc/sysconfig/icinga2

ICINGA2_RLIMIT_FILES=50000

Restart Icinga 2 afterwards, the systemd service file automatically puts the value into the application's environment where this is read on startup.

Setup Wizard Changes

Client and satellite setups previously had the example configuration in conf.d included by default. This caused trouble on config sync, or with locally executed checks generating wrong check results for command endpoint clients.

In v2.9.0 node wizard, node setup and the graphical Windows wizard will disable the inclusion by default. You can opt-out and explicitly enable it again if needed.

In addition to the default global zones global-templates and director-global, the setup wizards also offer to specify your own custom global zones and generate the required configuration automatically.

The setup wizards also use full qualified names for Zone and Endpoint object generation, either the default values (FQDN for clients) or the user supplied input. This removes the dependency on the NodeName and ZoneName constant and helps to immediately see the parent-child relationship. Those doing support will also see the benefit in production.

CLI Command Changes

The node setup parameter --master_host was deprecated and replaced with --parent_host. This parameter is now optional to allow connection-less client setups similar to the node wizard CLI command. The parent_zone parameter has been added to modify the parent zone name e.g. for client-to-satellite setups.

The api user command which was released in v2.8.2 turned out to cause huge problems with configuration validation, windows restarts and OpenSSL versions. It is therefore removed in 2.9, the password_hash attribute for the ApiUser object stays intact but has no effect. This is to ensure that clients don't break on upgrade. We will revise this feature in future development iterations.

Configuration Changes

The CORS attributes access_control_allow_credentials, access_control_allow_headers and access_control_allow_methods are now controlled by Icinga 2 and cannot be changed anymore.

Unique Generated Names

With the removal of RHEL 5 as supported platform, we can finally use real unique IDs. This is reflected in generating names for e.g. API stage names. Previously it was a handcrafted mix of local FQDN, timestamps and random numbers.

Custom Vars not updating

A rare issue preventing the custom vars of objects created prior to 2.9.0 being updated when changed may occur. To remedy this, truncate the customvar tables and restart Icinga 2. The following is an example of how to do this with mysql:

$ mysql -uroot -picinga icinga
MariaDB [icinga]> truncate icinga_customvariables;
Query OK, 0 rows affected (0.05 sec)
MariaDB [icinga]> truncate icinga_customvariablestatus;
Query OK, 0 rows affected (0.03 sec)
MariaDB [icinga]> exit
Bye
$ sudo systemctl restart icinga2

Custom vars should now stay up to date.

Upgrading to v2.8.2

With version 2.8.2 the location of settings formerly found in /etc/icinga2/init.conf has changed. They are now located in the sysconfig, /etc/sysconfig/icinga2 (RPM) or /etc/default/icinga2 (DPKG) on most systems. The init.conf file has been removed and its settings will be ignored. These changes are only relevant if you edited the init.conf. Below is a table displaying the new names for the affected settings.

Old init.conf New sysconfig/icinga2
RunAsUser ICINGA2_USER
RunAsGroup ICINGA2_GROUP
RLimitFiles ICINGA2_RLIMIT_FILES
RLimitProcesses ICINGA2_RLIMIT_PROCESSES
RLimitStack ICINGA2_RLIMIT_STACK

Upgrading to v2.8

DB IDO Schema Update to 2.8.0

There are additional indexes and schema fixes which require an update.

Please proceed here for MySQL or PostgreSQL.

Note

2.8.1.sql fixes a unique constraint problem with fresh 2.8.0 installations. You don't need this update if you are upgrading from an older version.

Changed Certificate Paths

The default certificate path was changed from /etc/icinga2/pki to /var/lib/icinga2/certs.

Old Path New Path
/etc/icinga2/pki/icinga2-agent1.localdomain.crt /var/lib/icinga2/certs/icinga2-agent1.localdomain.crt
/etc/icinga2/pki/icinga2-agent1.localdomain.key /var/lib/icinga2/certs/icinga2-agent1.localdomain.key
/etc/icinga2/pki/ca.crt /var/lib/icinga2/certs/ca.crt

This applies to Windows clients in the same way: %ProgramData%\etc\icinga2\pki was moved to %ProgramData%\var\lib\icinga2\certs.

Old Path New Path
%ProgramData%\etc\icinga2\pki\icinga2-agent1.localdomain.crt %ProgramData%\var\lib\icinga2\certs\icinga2-agent1.localdomain.crt
%ProgramData%\etc\icinga2\pki\icinga2-agent1.localdomain.key %ProgramData%\var\lib\icinga2\certs\icinga2-agent1.localdomain.key
%ProgramData%\etc\icinga2\pki\ca.crt %ProgramData%\var\lib\icinga2\certs\ca.crt

Note

The default expected path for client certificates is /var/lib/icinga2/certs/ + NodeName + {.crt,.key}. The NodeName constant is usually the FQDN and certificate common name (CN). Check the conventions section inside the Distributed Monitoring chapter.

The setup CLI commands and the default ApiListener configuration have been adjusted to these paths too.

The ApiListener object attributes cert_path, key_path and ca_path have been deprecated and removed from the example configuration.

Migration Path

Note

Icinga 2 automatically migrates the certificates to the new default location if they are configured and detected in /etc/icinga2/pki.

During startup, the migration kicks in and ensures to copy the certificates to the new location. This will also happen if someone updates the certificate files in /etc/icinga2/pki to ensure that the new certificate location always has the latest files.

This has been implemented in the Icinga 2 binary to ensure it works on both Linux/Unix and the Windows platform.

If you are not using the built-in CLI commands and setup wizards to deploy the client certificates, please ensure to update your deployment tools/scripts. This mainly affects

  • Puppet modules
  • Ansible playbooks
  • Chef cookbooks
  • Salt recipes
  • Custom scripts, e.g. Windows Powershell or self-made implementations

In order to support a smooth migration between versions older than 2.8 and future releases, the built-in certificate migration path is planned to exist as long as the deprecated ApiListener object attributes exist.

You are safe to use the existing configuration paths inside the api feature.

Example

Look at the following example taken from the Director Linux deployment script for clients.

  • Ensure that the default certificate path is changed from /etc/icinga2/pki to /var/lib/icinga2/certs.
-ICINGA2_SSL_DIR="${ICINGA2_CONF_DIR}/pki"
+ICINGA2_SSL_DIR="${ICINGA2_STATE_DIR}/lib/icinga2/certs"
  • Remove the ApiListener configuration attributes.
object ApiListener "api" {
-  cert_path = SysconfDir + "/icinga2/pki/${ICINGA2_NODENAME}.crt"
-  key_path = SysconfDir + "/icinga2/pki/${ICINGA2_NODENAME}.key"
-  ca_path = SysconfDir + "/icinga2/pki/ca.crt"
  accept_commands = true
  accept_config = true
}

Test the script with a fresh client installation before putting it into production.

Tip

Please support module and script developers in their migration. If you find any project which would require these changes, create an issue or a patchset in a PR and help them out. Thanks in advance!

On-Demand Signing and CA Proxy

Icinga 2 v2.8 supports the following features inside the cluster:

  • Forward signing requests from clients through a satellite instance to a signing master ("CA Proxy").
  • Signing requests without a ticket. The master instance allows to list and sign CSRs ("On-Demand Signing").

In order to use these features, all instances must be upgraded to v2.8.

More details in this chapter.

Windows Client

Windows versions older than Windows 10/Server 2016 require the Universal C Runtime for Windows.

Removed Bottom Up Client Mode

This client mode was deprecated in 2.6 and was removed in 2.8.

The node CLI command does not provide list or update-config anymore.

Note

The old migration guide can be found on GitHub.

The clients don't need to have a local conf.d directory included.

Icinga 2 continues to run with the generated and imported configuration. You are advised to migrate any existing configuration to the "top down" mode with the help of the Icinga Director or config management tools such as Puppet, Ansible, etc.

Removed Classic UI Config Package

The config meta package classicui-config and the configuration files have been removed. You need to manually configure this legacy interface. Create a backup of the configuration before upgrading and re-configure it afterwards.

Flapping Configuration

Icinga 2 v2.8 implements a new flapping detection algorithm which splits the threshold configuration into low and high settings.

flapping_threshold is deprecated and does not have any effect when flapping is enabled. Please remove flapping_threshold from your configuration. This attribute will be removed in v2.9.

Instead you need to use the flapping_threshold_low and flapping_threshold_high attributes. More details can be found here.

Deprecated Configuration Attributes

Object Attribute
ApiListener cert_path (migration happens)
ApiListener key_path (migration happens)
ApiListener ca_path (migration happens)
Host, Service flapping_threshold (has no effect)

Upgrading to v2.7

v2.7.0 provided new notification scripts and commands. Please ensure to update your configuration accordingly. An advisory has been published here.

In case are having troubles with OpenSSL 1.1.0 and the public CA certificates, please read this advisory and check the troubleshooting chapter.

If Icinga 2 fails to start with an empty reference to $ICINGA2_CACHE_DIR ensure to set it inside /etc/sysconfig/icinga2 (RHEL) or /etc/default/icinga2 (Debian).

RPM packages will put a file called /etc/sysconfig/icinga2.rpmnew if you have modified the original file.

Example on CentOS 7:

vim /etc/sysconfig/icinga2

ICINGA2_CACHE_DIR=/var/cache/icinga2

systemctl restart icinga2

Upgrading the MySQL database

If you want to upgrade an existing Icinga 2 instance, check the /usr/share/icinga2-ido-mysql/schema/upgrade directory for incremental schema upgrade file(s).

Note

If there isn't an upgrade file for your current version available, there's nothing to do.

Apply all database schema upgrade files incrementally.

# mysql -u root -p icinga < /usr/share/icinga2-ido-mysql/schema/upgrade/<version>.sql

The Icinga 2 DB IDO feature checks the required database schema version on startup and generates an log message if not satisfied.

Example: You are upgrading Icinga 2 from version 2.4.0 to 2.8.0. Look into the upgrade directory:

$ ls /usr/share/icinga2-ido-mysql/schema/upgrade/
2.0.2.sql 2.1.0.sql 2.2.0.sql 2.3.0.sql 2.4.0.sql 2.5.0.sql 2.6.0.sql 2.8.0.sql

There are two new upgrade files called 2.5.0.sql, 2.6.0.sql and 2.8.0.sql which must be applied incrementally to your IDO database.

mysql -u root -p icinga < /usr/share/icinga2-ido-mysql/schema/upgrade/2.5.0.sql
mysql -u root -p icinga < /usr/share/icinga2-ido-mysql/schema/upgrade/2.6.0.sql
mysql -u root -p icinga < /usr/share/icinga2-ido-mysql/schema/upgrade/2.8.0.sql

Upgrading the PostgreSQL database

If you want to upgrade an existing Icinga 2 instance, check the /usr/share/icinga2-ido-pgsql/schema/upgrade directory for incremental schema upgrade file(s).

Note

If there isn't an upgrade file for your current version available, there's nothing to do.

Apply all database schema upgrade files incrementally.

# export PGPASSWORD=icinga
# psql -U icinga -d icinga < /usr/share/icinga2-ido-pgsql/schema/upgrade/<version>.sql

The Icinga 2 DB IDO feature checks the required database schema version on startup and generates an log message if not satisfied.

Example: You are upgrading Icinga 2 from version 2.4.0 to 2.8.0. Look into the upgrade directory:

$ ls /usr/share/icinga2-ido-pgsql/schema/upgrade/
2.0.2.sql 2.1.0.sql 2.2.0.sql 2.3.0.sql 2.4.0.sql 2.5.0.sql 2.6.0.sql 2.8.0.sql

There are two new upgrade files called 2.5.0.sql, 2.6.0.sql and 2.8.0.sql which must be applied incrementally to your IDO database.

export PGPASSWORD=icinga
psql -U icinga -d icinga < /usr/share/icinga2-ido-pgsql/schema/upgrade/2.5.0.sql
psql -U icinga -d icinga < /usr/share/icinga2-ido-pgsql/schema/upgrade/2.6.0.sql
psql -U icinga -d icinga < /usr/share/icinga2-ido-pgsql/schema/upgrade/2.8.0.sql