I think, i know whats the problem. If you open the graph page a “FLUSH” command should be issued to the rrdcached daemon. Only after such a flush, you will see the data sent to the caching in the last minutes. With the default config rrdcached writes the data to disk after one hour or so. That is the reason why you see now older data.
You can check this behavior with the following command.
I can provide a little bit more insight as written before. With 2.4 the flush command for the single service looks to be missing.
With 2.3 this command is executed every time you open a page with graphs.
This looks to be missing with 2.4 and RAW edition. In enterprise edition i don’t see this problem.
In 2.3 the data is fetched with one “livestatus_lql” query from the “_fetch_rrd_data” function. Here it looks like that something inside livestatus was changed as it not automatically fetches the newest RRD data. Or better said the columns inside the livestatus table has only old data.
You don’t need to provide any further information as it is a internal problem how livestatus handle the requests for performance data.
As it is possible to reproduce the error, i think the rest needs to be done inside the CMK team.
yes, as Andreas says, there is nothing we need anymore. I was just looking for some other clues as I was assuming that we won’t be able to easily reproduce it. But we could reproduce it and will be working on a fix soon.
And thanks Andreas for your help and clarification.
Quick update from “behind the scenes”: the resolution will take a few weeks as the relevant Dev team is finishing another work first to keep the flow. But we will fix this before official release.
There is no release date, yet. Following the famous “It’s done, when it’s done”, we are working hard on testing, polishing and fixing to reach a high quality.