Stop logging NaN possible?

influxdb
#1

Is there a way to prevent this logging being send to the syslog server (not filtering at the syslog server)

 influxd: [I] 2018-01-19T21:23:56Z Dropping point ceph_value: NaN is an unsupported value for field value service=collectd
 influxd: [I] 2018-01-19T21:23:56Z Dropping point ceph_value: NaN is an unsupported value for field value service=collectd
 influxd: last message repeated 156 times
Error writing snapshot: compaction in progress: open /var/lib/influxdb/data/_internal/monitor/2/000000181-000000001.tsm.tmp: no such file or directory engine=tsm1
#2

Is anybody awake here? Or is influxdb maybe dying?

#3

Nobody dying here. NaN == Not a Number. So, you’ve got something in the ceph_value field which cannot be ingested into the database. I’d suggest having a look at your collector… and seeing what is being pumped into the collectd service endpoint?

#4

The problem is that collectd is a binary package so you are sort of at the hands of the developers to fix this. You should take such things into consideration when creating a ‘receiver’ for something like collectd.

Hence I think this is a valid request. I also put this as an issue on github where someone (developer?) agreed with this point of view.
Maybe introduce log levels or so, or move these to debug level. In any case, getting such/amount of messages in standard updated environment is not good.

#5

Log levels (along with a significant overhaul of logging in general) are being introduced with InfluxDB 1.5. You should be able to set your log level to WARN or greater to avoid sending this error on to your syslog service.

#6

You should have a proper look on how syslog works. Logging format is ‘not standard’.
Do you have an idea about how the production environment of your ‘professional’ clients looks like?