> do we have any existing node exporter which can read csv file?
No. But if you write a simple script to read the csv file and write it back
out in openmetrics format, then you can use node_exporter's textfile
collector. For example:
total_bytes{interface_name="abc"} 600
incoming_bytes{interface_name="abc"} 200
outgoing_bytes{interface_name="abc"} 400
error_bytes{interface_name="abc"} 10
invalid_bytes{interface_name="abc"} 0
rejected_bytes{interface_name="abc"} 20
But you should beware that:
1. Having different filenames for different times of day won't work. You
should create a single metrics file which has the *current* values, and let
prometheus scrape it periodically. It will timestamp the values with the
scrape time. In general, you cannot import historical data into prometheus
except as part of a one-time backfill operation.
2. You should not export the value which represents the counts from the
last 5 minute period. Rather, you should record monotonically incrementing
counters, i.e. the counter value which represents to *total* bytes since
the system started up or was last reset.
See: https://prometheus.io/docs/concepts/metric_types/
If your data source can only generate individual counts per 5-minute
period, then change your integration: use something like statsd_exporter to
maintain the counts, and push data to statsd_exporter every accounting
period. statsd_exporter will accumulate the values, and then get prometheus
to scrape that.
If you're not prepared to do that, then prometheus is probably not the
right system for you. Use a SQL database or similar instead.
> if not, any reference on how to write new node exporter?
Exporters simply respond to HTTP requests and return values in openmetrics
format. There are client libraries available in many languages.
https://prometheus.io/docs/instrumenting/clientlibs/
On Friday, 13 December 2024 at 08:50:16 UTC Navjyot Ramteke wrote:
> Hi All,
> We have a linux process which writes variables/statistics
> periodically into .csv file. We need to feed/save these
> variables/statistics in prometheus so that it can be viewed on Grafana.
> Our process periodically writes a new file in a folder(linux virtual
> machine).
>
> Folder:
> stats_13_12_12_00.csv >>>> written on 13th dec 12 am
> stats_13_12_12_05.csv
> stats_13_12_12_10.csv
> stats_13_12_12_15.csv
> stats_13_12_12_20.csv
> stats_13_12_12_25.csv
>
> in this example, the interval is 5 min. so next file will be written at
> 12:30 am.
> the contents of the file is given below:
> .csv file:
> 600,200,400,10,"abc",,20
>
>
> explanation:
> total_bytes=600
> incoming_bytes=200
> outgoing_bytes=400,
> error_bytes=10,
> interface_name="abc"
> invalid_bytes= <no value, so its not applicable>
> rejected_bytes = 20
>
> our csv file will just contains the values. the ordering of variables are
> fixed. if some variable/statistics is not applicable, then its value will
> be undefined (as shown below).
> our file will contain a single row. we will hav 3000-4000 such variables
> in each file.
>
> do we have any existing node exporter which can read csv file?
> if not, any reference on how to write new node exporter?
>
> Regards
> Navjyot.
>
>
>
--
You received this message because you are subscribed to the Google Groups
"Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion visit
https://groups.google.com/d/msgid/prometheus-users/9f1cfd65-8b66-490c-b4c5-8382b3151a4fn%40googlegroups.com.
