Werk #8353: Support for new RRD format with more efficient Disk IO

Component Metrics system
Title Support for new RRD format with more efficient Disk IO
Date Nov 12, 2015
Checkmk Edition Checkmk Enterprise (CEE)
Checkmk Version 1.2.7i4
Level Prominent Change
Class New Feature
Compatibility Compatible - no manual interaction needed

The Check_MK Micro Core now supports a new format when creating RRD files. Here:

LI:RRD files are kept below the new directory ~/var/check_mk/rrd. LI:All metrics of one service are kept in one RRD

Because now all metrics of a service are stored together in one RRD less disk blocks need to be updated when a new value is entered. This saves Disk IO. This is possible because the RRDTool version 1.5 now supports changing existing RRDs. That way if a check outputs new metrics these can be added to an existing RRD.

Switch to the new format

The new CMC format is not used automatically - neither for existing nor for new sites. The default for new sites might be changed in future, though. Switching your site to the new CMC format is done in the following steps:

  1. Make a backup of your current RRDs (~/var/pnp4nagios/perfdata). You might be tended to skip this step as the amount of data can be large. But you have been warned.

  2. Go to the ruleset Configuration of RRD databases of services (and hosts resp.)

  3. Create a rule for some (or all) hosts with the setting RRD storage format: One RRD per host/service (saves disk IO, only with CMC)

  4. Activate Changes - And now your system is in a state where new RRDs will automatically be created in the new format in ~/var/check_mk/rrd.

  5. In order to migrate the existing RRDs use the following command:

OMD[mysite]:~$ cmk --convert-rrds -v

It is also possible to specify a list of hosts and limit the conversion to these:

OMD[mysite]:~$ cmk --convert-rrds -v server01 server02

Note: This keeps the existing RRDs present in ~/var/pnp4nagios/perfdata and needs lots of diskspace. If you low in diskspace and are bold or if you have a backup (even better) you can directly delete the PNP format RRDs:

OMD[mysite]:~$ cmk --convert-rrds --delete-rrds -v

If you decided not to delete the old RRDs in the first place you can run the command with the option --delete-rrds at any time later.

Since all of you zigs of gigabytes of RRDs needs to be transformed this can take a long time. The good news: you do not need a downtime on your monitoring during this period. New RRDs are being used for storing current data as soon as they exist so with a bit luck and good disk performance you will not loose data points.

Old PNP SINGLE format

Note: The new RRD format is almost identical with the format SINGLE that was default in PNP4Nagios a long time ago. PNP was not able to alter the list of metrics that a service did output - however. So whenever the list of performance data of a check changed the graphs ended at that point of time. There is currently no direct migration from PNP's SINGLE format to the new CMC format. But you can go the intermediate step with cmk --convert-rrds --split.

To the list of all Werks