This message was deleted.

It should be fine mostly. Please use delta metrics. How many potential combinations of sale and payment do you expect? Are there any other attributes besides this?

There is around 200 sale and 200 payment so that would be 40,000 combinations I believe. Other attributes, yes, there would be 5 different applications with these attributes so we would also have service name attribute. So based on this is it still OK?

Yes, this is fine. While the total combinations can be 40,000, in practice it would be less than that I imagine with some combinations appearing more often than others. Regardless, it is still fine for counter metric. My recommendation is to use delta temporality for this metric because it performs relatively well. Also, how many CPU does ClickHouse has?

2 cpu's 4gb memory, is this enough?

I am assuming you will have 200k time series at max for this counter given you have 5 services. Might be okay depending on the number of queries you run against this data. What do you expect to do with this data? For example, what number of dashboards/panels/alerts that make of this counter?

we will have a dashboard per service(so 5) and each dashboard will have 1 panel where u select 1 sale and payment via variables to view the results over a period of time. We will have no alerts

It should be fine then. If you notice slowness, you may want to add more resources like 2 cpus for ingestion(including otel collector) and 2 cpus for querying etc...

ok thanks