Migrating existing data is not user friendly.
Your output of the “rrdtools info” command looks good. You have only one datasource (ds) and this datasource with 12 RRAs (RoundRobinArchive).
For your setup to store only the MAX values without consolidation you first need to define how-long do you want to store this data and with what time interval for one step.
The default value for 1 step is 1 minute.
One month without consolidation would need around 45000 datapoints. (60min * 24h * 31 days)
The running setup uses around 47000 datapoints to store data for 4 years.
Biggest difference here is that the data is stored in more than one RRA and every RRA has it’s own settings for consolidation.
From the output
rra[0].cf = "AVERAGE"
rra[0].rows = 2880
rra[0].cur_row = 2572
rra[0].pdp_per_row = 1 <-- 1 datapoint per step
rra[0].xff = 5.0000000000e-01
rra[0].cdp_prep[0].value = NaN
rra[0].cdp_prep[0].unknown_datapoints = 0
rra[1].cf = "AVERAGE"
rra[1].rows = 2880
rra[1].cur_row = 1788
rra[1].pdp_per_row = 5 <--- 1 datapoint every 5 steps (minutes)
rra[1].xff = 5.0000000000e-01
rra[1].cdp_prep[0].value = 0.0000000000e+00
rra[1].cdp_prep[0].unknown_datapoints = 0
Now the system takes 5 datapoints from rra[0] and consolidate it to one datapoint inside rra[1]. That is where you floating point numbers come from.
For the whole handling inside CMK you cannot manually create RRD files. These files are manage in case of the RAW edition by PNP4Nagios and inside the enterprise edition from the cmc core.