Monitoring: groups, new metrics and a longer history

Monitoring has received a larger update. Here's what changes.

Groups

• You can now bundle hosts into groups – for example “web servers”, “mail servers” or one per site.
• Define a check once at group level and it runs automatically on every host in the group. No more maintaining the same check on each host individually.
• Add a host to the group later and it picks up the group's checks on its own.
• Each instance still keeps its own status, its own history and its own alerts, so you can always see exactly which host is failing.

New metrics

• Availability, plus response times as median, p95 and p99.
• Incidents: open, resolved, mean time to recovery (MTTR), mean time between failures (MTBF) and longest single outage.
• Available at three levels: across your whole account, per host and per individual check.
• The time range is yours to choose, going back up to five years.
• Probes taken during an outage of my own infrastructure are excluded from your figures. A failure on my side won't spoil your statistics.

Tighter interval during an outage

• Each check can carry a second, shorter probe interval that kicks in as soon as the check goes red.
• Normal operation stays economical, while an outage is probed densely. You find out sooner when a service is back, without generating more load permanently.

A longer history

• Measurements are now compacted in stages instead of disappearing entirely after a fixed period: individual probes 7 days, per-minute values 90 days, hourly values 1 year, daily values 5 years.
• Previously the history ended after one year. You can now look back five years – finely resolved for recent data, coarser but longer-lived for everything older.
• Short ranges remain exactly as precise as before. For long ranges the analysis automatically reaches for the matching compaction level, which makes it considerably faster.

You'll find it all in the console under Monitoring, as usual. If something is missing or looks off, do send me a note.