Closed Friz-zy closed 7 months ago
Hello @Friz-zy A couple of questions of what you are trying to send. A very simple test
[[inputs.cpu]]
[[outputs.influxdb_v2]]
urls = ["http://localhost:3100/influx"]
worked. Can you share some information about what inputs you're using?
@akvlad you can check telegraf config here. This repo contain whole test setup
@Friz-zy I believe you need to use the influxdb_v2 output as @akvlad also suggested
@lmangani this problem appears just when influxdb_v2
used, which you can see in my first comment. Right now this section is comment out in the config as latest what I tested was vmagent from victoriametrics as a proxy between telegraf and qryn. I haven't tried outputs.influxdb
directly with qryn
@Friz-zy Couldn't reproduce your case locally with a small local setup. Will try your whole environment.
I'll send a couple of issues into your github repo as I can't make it start.
@akvlad I created a branch for you
git clone -b issues/462 git@github.com:Friz-zy/qryn-docker-s3.git /tmp/qryn
cd /tmp/qryn
docker-compose up -d
docker-compose ps
docker-compose logs -f telegraf
docker-compose logs -f qryn
docker-compose down
@Friz-zy thanks. Will check and fix the issue. Can't reproduce for now.
Hello @Friz-zy I have one suggestion for the error you experience. Please add a header to your configuration:
http_headers = {"Authorization" = "Basic dGVzdDp0ZXN0", "Content-Encoding" = "gzip"}
Is it possible that you send a gzipped body without the appropriate header? Please make sure that:
That is the only case when I see a similar error as you provide. On the other hand your telegram conf file provides no compression. It seems OK. Please try experimenting with the configuration:
## Content-Encoding for write request body, can be set to "gzip" to
## compress body or "identity" to apply no encoding.
# content_encoding = "gzip"
@Friz-zy fixed: https://github.com/Friz-zy/qryn-docker-s3/pull/1
@akvlad that works, thank you. But but I'm seeing some strange behavior in the metric names: when I use outputs.prometheus_client
or forward metrics via vmagent, then in tsdb metrics are stored in <module>_<metric name>
format, for example system_n_cpus
. But now with a direct entry in qryn I see a list of metrics without a module, for example n_cpus
. Thus a very important part of the metric name is eaten.... I pushed everything to main so you can check with grafana
@lmangani possibly need to update docs with examples how to connect collectors using login & password. Also I can update my fork and open a pull request to your original repo qryn-docker-s3
@Friz-zy As far as I understand module
you say about is saved inside measurement
tag value. Can you try reviewing the value of the measurement
field?
@akvlad yeah, measurement
has the right information. It's a little unusual but works, thank you
@Friz-zy It's about the format of influx -> prometheus conversion. We may reconsider the feature.
@akvlad yeah, I can totally picture it. After victoriametrics it is unusual for me, but ideologically this variant may be more correct. Although again telegraf itself generates metrics for Prometheus exactly in the <module>_<metric>
format...
Can this issue be closed?
I'm testing latest
qxip/qryn:bun
with telegraf. I have a problem with password authorization:When telegraf is configured with "Authorization" header in
outputs.influxdb_v2
qryn produce
metric parse error
Possibly this is a problem real metrics parsing because simple curl request works fine
When I tried url substitution in telegraf config
then telegraf just can't pass the authorization
Curl request works fine
Qryn is configured to use
test
as admin login and password