The default configuration does no scheduled device config backups, so the impact of DCB is negligible in cases where few or no backups are scheduled.
Still, documentation should mention that config backups share resources with the poller, and that it's a good idea to increase the poller's thread pool size if you expect to back up the configs of many devices.
This copy might take the form of an IMPORTANT: call-out, plus a short paragraph, inside the existing "Poller configuration" subsection before the "Configure TFTP Port" subsection.
- No scheduled backups will run in the default configuration, but if the user enables them, it's important to bear in mind the following points.
- Device config backups share resources with the poller, particularly poller threads
- If too many backups run simultaneously and take too long, the poller may not have sufficient free threads to do its other non-backup work. This could result in the poller failing to notice some service outages in your network.
- As a very rough rule of thumb, start with the number of devices you expect will ever be simultaneously backed up, multiply by the time (in seconds) you expect an average backup run to take, then divide by 3. Increase the existing thread pool size by the resulting figure.
- Changes to the poller's thread pool size take effect the next time OpenNMS is restarted.