Poller with multiple influxdb exporters forward only to the latest influxdb exporter from list

[figeac888]

Hello community,

We installed harvest-21.11.1-1.x86_64 on Red Hat Enterprise Linux release 8.4 (Ootpa). We configured 4 influxdb exporters and poller is only sending data to the latest influxdb exporter from list. Example /opt/harvest/harvest.yml :

Tools:
#  grafana_api_token: 'aaa-bbb-ccc-ddd'
#  autosupport_disabled: true

Exporters:
  influx_vdr1:
    exporter: InfluxDB
    addr: 10.114.176.27
    bucket: netapp
    org: socle
    token: XYXY
  influx_vdr2:
    exporter: InfluxDB
    addr: 10.114.176.28
    bucket: netapp
    org: socle
    token: XYXY
  influx_aub1:
    exporter: InfluxDB
    addr: 10.114.138.173
    bucket: netapp
    org: socle
    token: XYXY
  influx_aub2:
    exporter: InfluxDB
    addr: 10.114.138.174
    bucket: netapp
    org: socle
    token: XYXY

Defaults:
  collectors:
    - Zapi
    - ZapiPerf
  use_insecure_tls: false

Pollers:
  # this is a special/optional poller that provides stats about harvest itself
  # to remove this poller, simply delete the section "unix"
  unix:
    datacenter: local
    addr: localhost
    collectors:
      - Unix
    exporters:
      - influx_vdr1
      - influx_vdr2
      - influx_aub1
      - influx_aub2
  # add your Cdot / 7Mode pollers below, example:
  cluster-vdr:
    datacenter: XXXXXXX
    addr: 10.109.Y.X
    auth_style: basic_auth
    username: harvest
    password: XYXY
    use_insecure_tls: true   # since we don't use certificate authentication
    exporters:
      - influx_vdr1
      - influx_vdr2
      - influx_aub1
      - influx_aub2

Extract from /var/log/harvest/poller_unix.log :

{"level":"info","Poller":"unix","caller":"./poller.go:349","time":"2022-02-11T15:32:08+01:00","message":"poller start-up complete"}
{"level":"info","Poller":"unix","caller":"./poller.go:496","time":"2022-02-11T15:32:08+01:00","message":"updated status, up collectors: 1 (of 1), up exporters: 1 (of 1)"}

Extract from /var/log/harvest/poller_cluster-vdr.log

{"level":"info","Poller":"cluster-vdr","caller":"./poller.go:496","time":"2022-02-11T15:53:09+01:00","message":"updated status, up collectors: 32 (of 46), up exporters: 1 (of 1)"}

Harvest is only sending data to influx_aub2 for both pollers.

Regards, Kevin

[cgrinds]

hi Kevin - thanks for the feature request. You're right, currently Harvest only exports to a single influxdb per poller, last one wins.

Some folks have had success using Telegraf's InfluxDB listener to relay to multiple instance of InfluxDB. Not as convenient as having Harvest push to multiple instances but might help as a workaround.

[rahulguptajss]

Similiar request raised on discord

[rahulguptajss]

verified in main