Fixed
- Every BGP peer appeared twice on longer chart ranges, each copy cut short. A change in 1.79.0 started storing peer history under the connection name instead of the session name RouterOS reports, which is the right key — but the history already recorded under the old name stayed there, so a range covering both read as two different peers. The two halves are now joined back together on upgrade, and the older history is visible again in one line.
- Upgrading or rebooting a router raised a BGP alert for every peer on it. A reboot restarts all its sessions at once, which looked exactly like each session dropping on its own. On a rolling upgrade of five routers that was fifteen alerts in twenty minutes, and the same routers showed up as eleven "flapping" peers. A device that has just restarted is now silent on this alert — the reboot itself is already reported, once, instead of once per peer.
- The metrics-history check claimed hours of missing data on healthy devices. It counted how many readings arrived against how many a perfect five-minute rhythm would give, but collection runs a little slower than that in practice, so every device permanently looked 10-14% short and a full day reported over three hours lost. It now measures the longest continuous stretch with no reading in it, which is what a real gap is. The practical gain is the opposite of cosmetic: because the old count was always part-way to the alarm, a genuine gap shorter than about an hour and a half could never raise it.
Added
- The history check now also watches BGP collection, not just device health. The two are gathered separately, so BGP can stop being recorded while everything else keeps arriving — which is exactly what happened in the incident this check was built for. A device is flagged when either one has a hole, and the report says which.