Migrating to Version 1.1.9i1 and later

Migration to 1.2.0b1

SNMP Bulkwalk is now default

If you use WatoNG the first time a rule will be created which set bulkwalk as default for all SNMP Devices which not have the hosttag snmpv1

Migration to 1.1.13i3

Rittal CMC temperature check renamed

The service description of the check cmctc.temp has been renamed:

Check	Old description	New description
cmctc.temp	CMC Temperature %s	Temperature %s

If you want to keep the old name (because of continuity in performance graphs) then use the following code to your main.mk:

# Keep using legacy service descriptions
service_descriptions.update({
    "cmctc.temp"  : "CMC Temperature %s",
})

LSI MegaRAID physical disk check

The check megaraid_pdisks now handles multiple controllers installed in one server. The items that the check creates is no longer compatible with Check_MK versions before 1.1.8. Please reinventorize the affected hosts with:

user@host> cmk --checks=megaraid_pdisks -II YOURSERVERS...

Sounds in views enabled per default now

Alarm-sounds are now enabled per default in Multisite. If you do not like this, you can disable them with a new variable in multisite.mk:

enable_sounds = False

This can also be done with the new WATO. The variable sounds is now predefined with useful values.

Interface/port numbers are padded with zeroes

The checks if, if64, lnx_if and other compatible checks now pad port numbers with zeroes so that all ports of a host have the same display length and thus sort correctly. This means that in some cases your services will get other names when doing inventory the next time. Your historical performance data will not be continued in that case. You can switch back to the old behaviour with:

if_inventory_pad_portnumbers = False

Brocade port checks

The two checks fc_brocade_port and fc_brocade_port_detailed have finally been disabled and now always return UNKNOWN. Please use the new check brocade_fcport instead.

aironet_errors, statgrab_disk, statgrab_net: Performance data change

The three checks aironet_errors, statgrab_disk and statgrab_net now output rates in errors/sec or bytes/sec instead of raw counter values in their performance data. This change was neccessary in order to make Perf-O-Meters and NagVis Gadgets possible and to also give other addons direct access to the performance data. All other checks had already been converted previously or have been replaced by more modern variants.

For you this means: the existing RRD files of those checks are invalid. If you are using PNP4Nagios then you can rename the according RRD files and the matching XML files and let PNP create new files with the correct format. Via the PNP GUI you still have access to the renamed old RRDs. There is no way of continuing using the existing RRD (we are sorry).

Filesystem trend performance data

We needed to make another cleanup change to filesystem trending performance data. This is now normalized to MB / 24h. In prior versions if you changed the trend range then your historic data would be scaled wrong und thus be inconsistent. This problem is solved by the normalization.

This means - however - that all of your historic data that is using a range other than 24h (which is the default) will be scaled wrong for a last time. Note: this applies only to the growth and trend data, not to the actual filesystem usage.

Migration to 1.1.13i1

New WATO

This version comes with a complete overhauled and extended WATO. WATO now keeps all of its configuration files in check_mk/conf.d/wato. Hosts can be directly created in folders, the host lists have been removed. Hosts are now stored in the file hosts.mk within a folder.

If you want to migrate hosts from a 1.1.12-WATO, then you can use the script wato-migrate-1.2.0.sh. Call this script without any arguments while being in the same directory where main.mk is located. For OMD users this will be:

user@host> cd ~/etc/check_mk
user@host> ~/share/doc/check_mk/helpers/wato-migrate-1.2.0.sh

Please do not forget to make a backup of your configuration before using the migration script.

Migration to 1.1.12

Filter for service hard states

In order to avoid a naming conflict the names of the HTML variables used for the "Service hard states" filter have changed. If you have created custom views that use this filter, you have to update those views. The easiest way is to open the views in the view editor, check the settings of this filter and save the view again.

Migration to 1.1.11i3

Disk IO check

The new disk IO check, which has been introduced in 1.1.11i1, supports distinct monitoring for each device. In 1.1.11i1 and 1.1.11i2 this was the default behaviour. In situations where you have many filesystems (for example on larger Unix and Linux database servers), a huge number of new services would be created during inventory.

For that reason, we have set the default behaviour to creating one summary check per host. You can switch back to the old behaviour with:

diskstat_inventory_mode = "single"  # Default is now "summary"

If you are migrating for a version before 1.1.11 and want to keep the old behaviour of having one check for 'read' and one for 'write' then you should set the mode to legacy:

diskstat_inventory_mode = "legacy"  # Switch to very old behaviour

Window memory check

There is now a new dedicated check mem.win for Windows memory and page file usage. Inventory will automatically find and use the new check, which will have the service description Memory and page file and create a new RRD database. The check mem.used will not longer be found on windows systems. If you want to keep the current behaviour you can disable the new check:

ingored_checktypes += [ 'mem.win' ]

The inventory will not find mem.used on Windows systems even then, however.

Linux interface check lnx_if

The check lnx_if is now the default interface check for Linux hosts (requires an agent of at least 1.1.11i2). If you want to fall back to the old separate checks netctr.combined, netif.link and netif.params, then you can restore the old behavious in main.mk:

linux_nic_check = "legacy"

Direct RRD updates removed

The feature of direct RRD updates has completely been removed. The configuration variable do_rrd_update does not longer exist and must be removed from main.mk. If you have performance problems with too many RRD updates, please use the rrdcached instead (or use OMD which automatically sets up the RRD caching daemon).

Filesystem trend performance data

All filesystem checks (df, hr_fs, etc) now have performance data for trends enabled per default. This will produce two additional graphs per filesystem where you can see the current growth or shrinkage as well as the averaged growth over the configured time (default is 24 hours).

This means a change in the number of performance values. Your graphing addon might run into trouble if this is the case. If you use PNP4Nagios and have set RRD_STORAGE_TYPE set to SINGLE in process_perfdata.cfg (still the default), then your graphs will no longer be updated. Please either convert your graphs to storage type MULTIPLE, or delete your RRD graphs of your filesystems or switch back to the old behaviour:

# Switch back default to pre-1.1.11i3:
filesystem_default_levels["trend_perfdata"] = False

Some temperature services have been renamed

In order to make configuration easier und the monitoring data more consistent, the service descriptions of some checks have been renamed (we apologize for not having chosen consistant descriptions in the first place). This applies for the following checks:

Check	Old description	New description
decru_temps	TEMP %s	Temperature %s
f5_bigip_temp	TEMP %s	Temperature %s
fsc_temp	FSC TEMP %s	Temperature %s
hp_proliant_temp	HW Temp%s	Temperature %s
akcp_sensor_temp	Temperature Sensor - %s	Temperature %s
apc_symmetra_ext_temp	APC External Temp %s	Temperature External %s
apc_symmetra_temp	%s	Temperature %s
smart.temp	SMART %s Temperature	Temperature SMART %s
nvidia.temp	NVIDIA Temperature %s	Temperature NVIDIA %s

Note: From Nagios' point of view the renamed services are new services with their own new history. New RRDs will be created. You will not be able to read a contigous history of those services from the log files. If you do not like this, you can either try to rename the entries in the logfiles and rename your RRDs, or you can tell Check_MK to still use the old service descriptions by adding the following code to your main.mk:

# Keep using legacy service descriptions
service_descriptions.update({
    "decru_temps"           : "TEMP %s",
    "f5_bigip_temp"         : "TEMP %s",
    "fsc_temp"              : "FSC TEMP %s",
    "hp_proliant_temp"      : "HW Temp%s",
    "akcp_sensor_temp"      : "Temperature Sensor - %s",
    "apc_symmetra_ext_temp" : "APC External Temp %s",
    "apc_symmetra_temp"     : "%s",
    "smart.temp"            : "SMART %s Temperature",
    "nvidia.temp"           : "NVIDIA Temperature %s",
})

Migration to 1.1.11i2

WATO

WATO now allows to configure the host files via the GUI. The variable config_files in multisite.mk is obsolete and will not be used by WATO anymore.

If you have already used WATO, then this is the way for migrating your existing hosts into the new WATO:

• Move the existing WATO-handled configuration files in conf.d away
• Go to the WATO main directory and re-create those files via the GUI with the same names
• Move your WATO files back into place

Now your hosts should appear again.

Brocade switch ports

In previous versions there were two variants of checks for Brocade FC switch ports: fc_brocade_port and fc_brocade_port_detailed. These two checks are new being replaced with the new check brocade_fcport. The previous checks still work but are deprecated and not longer found by the SNMP scan. Please think about reinventorizing the hosts and questions in order to use the new checks.

In order for PNP4Nagios to be able to create RRD files, you need to delete (or move away) your existing RRD files for the switch port checks.

snmp_info and snmp_uptime

These two checks now will be automatically added when doing an inventory (in WATO or with cmk -I or -II). If you do not need those checks, please use ignored_checktypes or ignored_checks. For details refer to the section about excluding checks from inventory. The following configuration code will globally disable inventory of these checks:

ignored_checktypes += [ 'snmp_uptime', 'snmp_info' ]

Livestatus: UTF-8 encoding

When Livestatus outputs data in JSON or Python format, it encodes non-ASCII characters with escape sequences. In versions before 1.1.11i2 2-Byte UTF-8 sequences in the Nagios data (the object configuration files) where recognized and properly encoded. Other non-ASCII characters where assumed to be encoded in Latin-1. That way, western special characters and many other special characters worked, but for example not the €-Symbol and also not Chineese.

The new version of Livestatus calls this mode mixed and now offers also the modes utf8 and latin1. Important: The new default is utf8. If (and only if) your input files are Latin-1 encoded, then you need to add the option data_encoding=latin1 to your line where the Livestatus module is loaded (int nagios.cfg). If you want to keep the old mode, set data_encoding=mixed. It is very probable - though - that you do not need any change in your configuration.

For details please refer to the Livestatus documentation.

Migration to 1.1.11i1

Dictionary based configuration for df and friends

All filesystem checks have been unified. Currently these are df, df_netapp, df_netapp32, hr_fs and vms_df. They now use the same implementation and share all features. Also there is a new and alternative way of specifying check parameters: a dictionary instead of a tuple or triple. When using filesystem_default_levels, you have to use the new dictionary based method. If your levels are (90, 95) and you do not use the magic number, then please convert your main.mk to:

filesystem_default_levels = {
    "levels" : (90, 95)
}

With using the magic number this is:

filesystem_default_levels = {
    "levels" : (90, 95),
    "magic"  : 0.6,
}

The same with enabling performance data for trends:

filesystem_default_levels = {
    "levels"         : (90, 95),
    "magic"          : 0.6,
    "trend_perfdata" : True,
}

The old tuple syntax like (90, 95) or (90, 95, 0.6) is still allowed in check_parameters.

Migration to 1.1.10

For the first time in the latter history of Check_MK, we decicded to make some incompatible changes to Check_MK. This means that you might need to adapt your configuration a bit. The changes have been neccessary to overcome some architectural problems and to clean up and straighten things a bit (making them easier).

This article tells you all you need to know about theses changes.

Inventory, tcp and snmp

Hosts might be contacted via TCP (Check_MK agent on port 6556), SNMP both ways or none of these. This now must be configured in all_hosts via special host tags. Please use the following tags:

all_hosts += [
  "server01",               # this host uses only TCP
  "server02|tcp",           # this host uses only TCP
  "switch01|snmp",          # this host uses only SNMP
  "windows01|tcp|snmp",     # this host uses TCP and SNMP
  "router01|ping",          # this host is only pinged
]

As you can see from this example:

• Hosts using only the Check_MK agent may be tagged with tcp, but need not.
• Hosts using only SNMP must be tagged with snmp. In 1.1.8 this was proposed. Now it is obligatory.
• Hosts using SNMP and TCP need to be tagged with tcp and snmp. This is new.
• Hosts that should be only pinged should now be tagged with ping. This will prevent inventory from trying to contact the agent or use SNMP. As long as you do no inventory on those hosts, nothing dangerous will happen if you leave out the tag, though.

Calling inventory

The syntax for calling inventory has simplyfied. Now you omit the tcp and snmp and simply write:

root@linux# check_mk -I server01

The host tags ping, tcp and snmp now control, which check types are being tried to inventorize. If you want to specify the check types to inventorize manually, use the new option --checks:

root@linux# check_mk --checks snmp_info,if64 -I switch01 switch02

Config-Check

As you might know check_mk -X checks your configuration for correct python syntax and misspelled variable. Now this is done everytime you call check_mk. So you really need to fix your configuration before calling any command. This saves you against undetected misspelled variables.

non_bulkwalk_hosts

This variable has been dropped. Please specify bulkwalk_hosts positively now. You can use NEGATE in order to configure exceptions:

bulkwalk_hosts = [
 ( NEGATE, [ "oldstuff" ], ALL_HOSTS ), # Oldstuff devices use SNMP V1
 ( ALL_HOSTS, ),                        # all other devices use SNMP V2C
]

Setup, Multisite and url_prefix

Setup and Multisite (in sites) now use url_prefix instead of pnp_url, nagios_url and nagios_cgi_url. In most cases the url_prefix is /. In OMD the url_prefix contains the site, e.g. /mysite/. For PNP4Nagios a pnp4nagios/ is appended automatically, for Nagios a nagios/ or a nagios/cgi-bin/. The url_prefix is also used for access to the Logwatch pages and to WATO.

What do you need to do? Update your sites in multisite.mk, if you are using a multi site setup:

sites = {
   # connect to local Nagios
   "local" : {
        "alias"      : "Munich",
        "url_prefix" : "/muc/"
   },

   # connect to remote site
   "paris": {
        "alias":          "Paris",
        "socket":         "tcp:localhost:6555",
        "url_prefix":     "http://srvpar01/paris/",
    },
}

If you want a HTTP connection to your remote addons that is totally transparent for the user, please also read our article about using mod_proxy.

Check type kernel

The check type kernel is now not longer using the internal counter name as check item but a description. So the service Kernel ctxt has now vanished and a new service Kernel context switches will apear at inventory. Also the new check has a new performance data format which allows to display a perf-o-meter in multisite.

Your task: re-inventorize the kernel checks on all Linux hosts.

filesystem_levels

The configuration parameter filesystem_levels has finally been removed. You can do the same thing with the new and more general paramter check_parameters. The difference is only slight. check_parameters always refers to the service description (which always starts with fs_ for filesystems). filesystem_levels referred to the name of the filesystem.

Livestatus and Multisite

The new version of Multisite uses some new features that have been introduced in the new version of Livestatus. Please make sure that you update all you Nagios sites to at least the same version as Multisite.