Kapacitor: 1.3.1
Influxdb 1.0.0rc1
I’m finding that Kapacitor is extremely inconsistent with its behavior - sometimes it works, sometimes it doesn’t seem to be working, I poke around, reload the tasks, restart, and then some of the alerts start working.
I’m now reasonably convinced that the issues are somehow related to the subscriptions in influxdb, so a few questions:
- Are there supposed to be kapacitor subscriptions to “name: _internal”, retention policy ‘monitor’?
- Can I safely delete these? I have found some bug reports that this corrupted the influxdb, and this would be extremely painful for me…
Possible clue - In the influxdb log messages can see kapacitor very frequently querying influxdb for its subscriptions.
If you delete the subscriptions nothing bad will happen, but Kapacitor will notice they are missing and add them back in.
(hence the periodic checking of subscriptions). If you want to permanently remove them see the [[influxdb]] excluded-subscriptions configuration section. This will tell Kapacitor to not subscribe to those listed databases.
As for the stability what protocol are you using for scubscriptions UDP or HTTP?
If you are using UDP then I would recommend switching to HTTP to avoid a whole class of transport issues.
If you are using HTTP do you see error in the InfluxDB log about not being able to write to Kapacitor? If so what are the errors?
Hi,
I am facing similar issues with subscriptions. I use these to get stats information and raise deadman alerts. My system consists of:
- machine1: Linux machine with influxdb
- machine2: Linux machine with kapacitor and influxdb
I have all configured properly, influxdb but start not allways works, even measurements in machine1 has data, the alerts raised.
I check subscriptions in machine1 regularly and are pointed to localhost http://localhost:9092 which is not correct. The correct is http://machine2:9092.
Any way to check subscription process to find out the issues?
Thank you
