NMIS collects all performance in 5 minute intervals by default, but doesn't keep the raw/most precise data forever - disks are big but not infinite.
The summarisation you are complaining about is a feature of RRDtool, the round robin database system that nmis uses: it guarantees that the rrd files always stay at the same size.
naturally you can adjust how long you'd like NMIS to keep the precise data, and that is documented extensively on the wiki page "Amount of Performance Data Storage NMIS8 Stores".
Hi Alex
Thanks for your response. We changed the file content as per your link under Common-database.nmis, however It didnt make any difference. Its been over 2 weeks since we changed the values but if I export a report which is over 8 days it's compressing to 30 min. rather than expected 5 min intervals.
Do we have to change any thing else or restart a service!
That's how our Interface section looks like under /usr/local/nmis8/models/Common-database.nmis
'interface' => {
'rows_day' => '9216',
'rows_month' => '2268',
'rows_week' => '4608',
'rows_year' => '1890',
'step_day' => '1',
'step_month' => '24',
'step_week' => '6',
'step_year' => '288'
Keep (days) | Summarise (minutes) | Model Entry | Default Value | New Value |
---|---|---|---|---|
32 days | 5 minutes | rows_day | 2304 | 9216 |
96 days | 30 minutes | rows_week | 1536 | 4608 |
Thanks
the rrd size change does not apply to existing files. it is possible but not trivial to convince rrdtool to apply them to already stored data; this is discussed in detail here: See the section on Migration of existing data to a new RRD size scheme.