Reading logs from a several servers #270
-
|
Is your feature request related to a problem? Please describe. Hi, I'm discovering the possibility to run the exporter on a separate server and collect the Nginx logs from several servers. I'm not sure about the Describe the solution you'd like A possibility to configure several servers where to collect Nginx logs. Describe alternatives you've considered Reading from syslog? Installing the exporter on each server? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 3 replies
-
|
Hi @martin-helmich! |
Beta Was this translation helpful? Give feedback.
-
|
Hey Andrii; the exporter is not designed to scrape metrics from different servers. If you want to aggregate metrics from multiple servers, the generally accepted architecture for this in the Prometheus ecosystem is to have an exporter running on each server and have them scraped by a single, central Prometheus instance. Prometheus will store each of those as a separate time series (typically differentiated by a If you really want, the syslog approach might work, although untested. You can configure the listen address with the Hope that helps! |
Beta Was this translation helpful? Give feedback.
-
|
@martin-helmich thank you for the response! |
Beta Was this translation helpful? Give feedback.
-
|
@andrii-bodnar / @martin-helmich - I had a similar use-case wherein I wanted to setup a central nginxlog exporter which will accept logs from a fleet of nginx servers. Essentially a remote syslog server which will accept data and nginxlog exporter will export that data. I wanted to know, were you able to set it up successfully? Any challenges so far that you have encountered? |
Beta Was this translation helpful? Give feedback.
-
|
Hi @bs-amitmahajan29, I have abandoned this idea and ended up with the exporter per server setup. |
Beta Was this translation helpful? Give feedback.
Hey Andrii; the exporter is not designed to scrape metrics from different servers. If you want to aggregate metrics from multiple servers, the generally accepted architecture for this in the Prometheus ecosystem is to have an exporter running on each server and have them scraped by a single, central Prometheus instance. Prometheus will store each of those as a separate time series (typically differentiated by a
joblabel), but you can easily aggregate using PromQL aggregation functions.If you really want, the syslog approach might work, although untested. You can configure the listen address with the
.namespaces[*].source.syslog.listen_addressproperty. However, you'd probably need to fi…