Impact of connectivity issues on distributed polling?

We’re looking at potentially using LibreNMS across multiple offices. They’re linked via site to site VPNs so in theory we can set up multiple pollers at each office as described in https://docs.librenms.org/Extensions/Distributed-Poller/ however I am wondering what happens if a VPN link is down for some time period (seconds, minutes, …) and an office can’t reach whichever one is running rrdcached ?

Is there some mechanism that will buffer the updates from the disconnected poller and then retroactively fill in the data, or will the data be lost? Will a loss of connectivity crash the poller?

If the only way to reliably not lose polling data is to use a separate instance for each site, then we might do that, but if we can use the distributed model and not have issues if a VPN goes down for some reason, that would be ideal. Failing that, is there any existing way to easily link multiple separate instances together, to avoid multiple logins / browser tabs ?

1 Like

Distributed polling use central rrdcached to whom all poller feed data, so if there is no link you will have gaps for poller group which missed link to rrdcached. I don’t know about any buffering possiblities here, so probably you better go for vpn redundancy to keep link between pollers and central server or use smaller separate instances on each site.