Hello,
I’m having an issue with correctly calibrating my grafana alarms. We use discord hooks to get notified.
This is they query I’m using:
(node_memory_MemTotal - node_memory_MemAvailable)/node_memory_MemTotal
I’m applying classic_conditions avg above 0.85.
This works as expected, in that, if several instances go above 0.85, only one capture is being sent (and this is how expect it to work). This is how the alarm is configured:
If, on the other hand, I use the modern way Reduce - Mean (Strict) + Threshold is above 0.85 - I get a screen capture for each instance that exceeds 0.85. Which is redundant and not desirable and it’s hard to understand what is what.
I’d like the alarm to be kept as succinct as possible in this case and just to know that at least one instance had exceeded that threshold, so just one capture.
- Any ideas how I can do that in the modern way?
- Have the classical conditions become obsolete? Are we supposed to avoid them in favour of the new way?
I guess what I’m asking is, can you reduce a multi-dimensional alarm to one dimension in any other way than using classical conditions? Or should I just use the classical condition and just get on with it?