Send metrics to nat with timed out errors.

[ozgurerdogan]

I am running influxdb in a nat of pve kvm environment.

I see randomly:

metrics send error 'Monitor': 500 Can't connect to 8.8.8.8:8086 (Connection timed out) but metrics are being collected.

I have forwarded 8086 to influxdb internal nat ip (in pve).

What could be issue?

 

[ozgurerdogan]

I have tried dedicated vds with public ip but I still see these timeouts. So I guess it is not on remote site (influxdb) but pve side. So I think I can ignore them. But they cause small short outages on pve graphs.

 

[sb-jw]

So basically it works, did I understand that correctly?

Are you sure it's because of the NAT? Maybe an InfluxDB is simply not configured optimally and the VM / container has too few resources? How big is your setup that writes there?

 

[ozgurerdogan]

Yes it works. I also tried without nat. Well I am trying with more resources... Does influxdb really need fast hardwares?

 

[sb-jw]

Does influxdb really need fast hardwares?

It depends, I gave my Go-Graphite instance 8 cores, 16 GB of RAM and a RAID10 made up of 6 SSDs so that the metrics can be written accordingly. The instance must be able to process 10 MB/s for me.

That's why I asked about the size of your cluster. Otherwise you still have all your logs and certainly also monitoring in which you can see the utilization.

 

[ozgurerdogan]

You might be right but even with one node, I see these timeouts. Also in forum, there are other users..