You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have installed the nuget and I get metrics at https://example.com/metrics, but since I have multiple web server instances, in every Prometheus request the load balancer respond the metrics from different instance every time.
That way my counters have a portion of the counter from one instance and a portion of the other instance.
They should be either
two different scrap jobs, or
combined together
To do the first I need to create two jobs in Prometheus config that access the two instances independently, and then join them in query.
Unfortunately, this is not possible in Azure Web Apps because to access a specific instance the request must set a cookie, and prometheus doesn't support sending the cookie 'ARRAffinity=....' (only auth cookies).
I wish if there was a way to do the second. Maybe with using a Redis type of cache. That way if two processes are setup to use the same Redis server and the same Key, they would increase the same counter and their /metrics would return the same data.
The text was updated successfully, but these errors were encountered:
I have installed the nuget and I get metrics at
https://example.com/metrics
, but since I have multiple web server instances, in every Prometheus request the load balancer respond the metrics from different instance every time.That way my counters have a portion of the counter from one instance and a portion of the other instance.
They should be either
To do the first I need to create two jobs in Prometheus config that access the two instances independently, and then join them in query.
Unfortunately, this is not possible in Azure Web Apps because to access a specific instance the request must set a cookie, and prometheus doesn't support sending the cookie 'ARRAffinity=....' (only auth cookies).
I wish if there was a way to do the second. Maybe with using a Redis type of cache. That way if two processes are setup to use the same Redis server and the same Key, they would increase the same counter and their
/metrics
would return the same data.The text was updated successfully, but these errors were encountered: