Distributed polling speed

Hi Guys,

Firstly sorry if this has come up before I did look around and found similar topics but nothing as specific.

So I’ve just finished setting up distributed polling with my central box in the UK and the poller in Asia hitting local devices. I naively hoped that this would resolve my long polling times as my checks were running across an internet VPN.

Instead they remain pretty much the same, I’ve moved the latency issue away from the snmp GET/Walk to the SQL functions.

Poller results run from the UK:
#### End Alerts ####
SNMP: Get[21/6.00s] Walk [27/168.31s]
MySQL: Cell[44/0.02s] Row[2/-0.01s] Rows[20/0.02s] Column[3/0.00s] Update[283/0.08s] Insert[2/0.00s] Delete[0/0.00s]

Poller results run from the Asia poller:
#### End Alerts ####
SNMP: Get[21/0.44s] Walk [27/32.98s]
MySQL: Cell[44/34.39s] Row[3/-22.14s] Rows[21/9.23s] Column[3/1.30s] Update[283/73.94s] Insert[2/1.56s] Delete[0/0.00s]

Basically I’m now stuck for ideas on improving this as disabling poller modules or segregating modules out to more pollers wont solve the distance issue.

I’m not sure what we can suggest to improve this. You’ll also have the latency on the rrdcached queries as well.

Thanks for your response, I thought that might be the case.

Luckily, our Asia sites aren’t as big, so we may be able to just live with the longer poll times.

I wasn’t sure about something like master-master DB replication between both sites, but I pretty much never work with databases so I wasn’t sure on any implications there. And again that would just be moving the latency delay elsewhere.

It may help a little but not every part of LibreNMS is cluster aware so you might then have other issues. Latency for the DB replication will still be there as will the latency for rrd updates :frowning: