We have a series of alerts setup in kapacitor, which we've been using fine for years. After upgrading to InfluxDB OSS v1.11.7, we've seen a few things broken in kapacitor (using Kapacitor OSS v1.7.5 on Debian):
- deadman alerts started firing, because no data arrived
- the kapacitor log was spammed with messages such as
Oct 29 14:23:00 [redacted] kapacitord[11831]: ts=2024-10-29T14:23:00.613Z lvl=error msg="failed to parse points" service=influxdb cluster=localhost listener_id=udp:[redacted].default err="unable to parse '[redacted]': missing fields"
We use a setup with kapacitor subscribing via udp to instance on the same server. My suspicion is something isn't working properly anymore - this may be an Influx issue? Or is kapacitor using some kind of legacy way to subscribe to v1 instances that is no longer functional?
I did try switching to http subscriptions which fixes the log spam, but does not fix data being seen by kapacitor.
In the meantime we'll probably downgrade to InfluxDB OSS v1.8.10 again.
We have a series of alerts setup in kapacitor, which we've been using fine for years. After upgrading to InfluxDB OSS v1.11.7, we've seen a few things broken in kapacitor (using Kapacitor OSS v1.7.5 on Debian):
Oct 29 14:23:00 [redacted] kapacitord[11831]: ts=2024-10-29T14:23:00.613Z lvl=error msg="failed to parse points" service=influxdb cluster=localhost listener_id=udp:[redacted].default err="unable to parse '[redacted]': missing fields"We use a setup with kapacitor subscribing via udp to instance on the same server. My suspicion is something isn't working properly anymore - this may be an Influx issue? Or is kapacitor using some kind of legacy way to subscribe to v1 instances that is no longer functional?
I did try switching to http subscriptions which fixes the log spam, but does not fix data being seen by kapacitor.
In the meantime we'll probably downgrade to InfluxDB OSS v1.8.10 again.