Add example of translating new metrics to old format in case of migration to 1.16 version#982
Conversation
|
Oh well, I think I got your idea. These rules are for translating old format to the new one. So you wanted users to make all the metrics in the new format and redo all the Grafana dashboards, right? |
|
So, I've added new example that is basically just does opposite things then yours. Motivation: I want to keep all the software I run in infrastructure up-to-date. But I don't want to redo hundreds of dashboards because of upgrading node_exporter. I know that it should happen one day, but, for example, not today :P |
|
Hi, thanks! You'd need to sign you commits, see https://github.com/prometheus/node_exporter/blob/master/CONTRIBUTING.md#contributing In general not sure we need this here but I guess if it's useful there isn't harm. @SuperQ thoughts? |
|
I'm not sure we really need this, but I don't see any harm in providing the sample. We (GitLab) didn't go this route for our Grafana, we did a sweep of dashboards and added both new and old metric queries to the dashboards, this was far more seamless and doesn't have the duplicated storage requirements. |
|
Well, it's a nice thing to do, of-course, to change everything to the new format at the very first place, but when you really have hundreds of dashboards, custom alerts and recording rules - it's a task that would take a while, day or couple for anyone, and it's quite expensive for just upgrading node_exporter version and it's the time you may probably not have right now. So, if you guys think it's not the case - feel free to close the PR, since I've already applied it in our infrastructure, but I think someone like me would really thank you for adding such a doc there. |
99b913b to
e36aa44
Compare
Signed-off-by: Ivan Kiselev <ivan@messagebird.com>
Signed-off-by: Ivan Kiselev <ivan@messagebird.com>
Signed-off-by: Ivan Kiselev <ivan@messagebird.com>
|
sorry for the delay guys, all the commits are signed now. |
…tion to 1.16 version (prometheus#982) Add additional example of how to save old metrics Signed-off-by: Ivan Kiselev <ivan@messagebird.com>
Hello people!
Seems like there's a misconfiguration in the rules for the backward compatibility between node_exporters < 1.16 and 1.16.
As said in the changelog -
node_cpu_seconds_totalis a new metric name, and thenode_cpuis the old one, so they should be put vice-versa.