Regions and probes

Learn how InfraNest checks your service from multiple locations around the world, and what that means for alerts, uptime numbers, and the charts you see on a monitor's page.

#Overview

• InfraNest runs your checks from several locations ("probes") so a problem in one place doesn't trigger a false alarm, and you can see how your service performs from different regions.
• A failure seen from any one region counts as real — a firewall that blocks one part of the world is an outage for the visitors there.
• You can let InfraNest pick the best location automatically, choose your own preferred regions, and compare how regions perform against each other right on the chart.

#How regions and consensus work

1. Hover a block in the latest checks strip on a monitor's overview to see which region ran that check and when.
2. Open the chart on a check's detail page and choose Compare regions to see response times side by side.
3. If an alert email arrives, check it — it lists which regions saw the site down and which still saw it up.
4. Check the Failed checks before alerting setting on your monitor. This decides how many checks in a row must fail (2 by default) before an alert fires, and it stops a single blip from paging you. Each region counts its own failures separately.
5. Remember that slow responses show as degraded rather than down.
6. If two of our probes in the same region disagree, the failure isn't counted — that's a problem with our machine or its route, not your site.
7. If a check says your site is down but it loads fine for you, see the article A monitor says my site is down, but it loads for me.

NoteA location that can't reach your site at all on the first try gets one immediate retry. Only two failures in a row from that location count as down. A site that answers with an error or too slowly is never retried — that result stands.

You can also pin preferred regions and choose IPv4 or IPv6 per check. The default, prefer IPv4, uses IPv4 whenever your site has an IPv4 address, and only falls back to IPv6 if it doesn't.

#When our probe is the problem

Once in a while, one of our own probes has trouble — it can't reach part of the internet, or it starts failing many unrelated sites at once. InfraNest watches for this automatically.

1. While a probe is having trouble, every failed check it makes is marked Likely not you.
2. Any incident that no other location confirmed is marked the same way.
3. Look for this label on the monitor's page, on the incident, and in the alert itself.
4. For more detail, see When it's likely not you in the Incidents article.

#Review suggestions about where a monitor checks from

InfraNest reviews what each location has actually seen and may suggest a change under Watch better on the Monitoring page. Nothing changes automatically — you choose whether to apply it.

1. Go to the Monitoring page and look for a suggestion under Watch better.
2. Read the suggestion type:
   • "These regions have never disagreed" — shown when, for at least two weeks, every location saw the same result at the same time. Applying it keeps only the two nearest locations and tells you how many checks an hour you'll save. It never drops below two locations, so a probe issue can still be told apart from a real outage.
   • "Fails from one region only" — shown when one location repeatedly sees your site down while others see it up. The suggestion names the days this happened and offers to add the nearest unused location for comparison, or — if you already use every location your plan offers — to wait until two regions agree before alerting.
3. Click the suggestion and confirm it to apply the change.
4. If you change your mind later, update the locations again in the monitor's settings.

NoteSuggestions only appear once there's enough history — a very new or rarely-checked monitor won't have any yet, and suggestions only ever include locations your plan supports.

#Understand monitors checked from one location

A monitor checked from a single location shows one line under its result, such as Checked from Amsterdam, instead of a card per location. There's no Compare regions option and no per-location lane on the Over time card, since there's nothing to compare.

1. If this monitor's check fails, wait a moment — InfraNest automatically asks another probe for a second opinion before counting the failure.
2. If the second probe also fails to reach your site, the failure counts as usual. The page reads, for example, Down from Amsterdam, confirmed from Frankfurt.
3. If the second probe succeeds, the failure isn't counted, nobody is alerted, and your uptime is unaffected. The page reads Didn't reach it from Amsterdam; Frankfurt could, so this wasn't counted.

This extra check only happens when a check fails, and only costs one additional check. If the second probe doesn't respond within a few minutes, the original failure counts as it normally would.

If you leave the monitor's location set to Auto (the default for new monitors), or your plan only checks from one location, InfraNest picks the best location for you:

1. InfraNest starts with the region nearest to your site.
2. It runs one check from each region your plan allows.
3. The fastest region keeps the monitor. The line then reads Checked from Amsterdam · picked automatically, closest to your site.
4. InfraNest re-measures about once a week, moving your monitor only if another region is clearly faster, or immediately if the current region's probes become unavailable. A move appears on the Over time card as Now checked from … instead of ….
5. To stop automatic moves, pick a location yourself in the monitor's settings.

Owners and admins on plans that check from only one location will also see Check from more locations, which shows which plans include checking from several places — useful for telling an outage of your own apart from a network problem affecting just one region.

#Set the alert condition: any region or all regions Business

The Alert condition on a check decides how many regions must report a problem before you're alerted.

1. Open the check's settings and find Alert condition.
2. Choose Alert when any region reports down (the default) if every region matters to you — for example, if you have customers in each one. One region failing is enough to alert you.
3. Choose Alert when all selected regions simultaneously report down if you want to be alerted only when every picked region agrees your site is down at the same time. Pick this if one region is known to be unreliable and you don't want it alerting you on its own.
4. Remember that Failed checks before alerting still applies on top of this condition — the condition must hold for that many checks in a row before you're alerted.

With any, the alert ends once the failing region recovers — a region that was fine the whole time doesn't end it. With all, the alert ends the moment "every region is down" stops being true.

NoteIf a region you picked has no measurement for a while (its probe is offline), it can't vote either way — the regions still reporting decide. This way an offline probe never silently turns your alerts off. With only one region selected, both settings behave the same.

#Read the uptime bars and percentage

• Green means no confirmed downtime in that period. A single location failing a check now and then, without other locations agreeing, doesn't count as an outage — hover the bar to see how many checks failed without being confirmed.
• Amber means part of the period was confirmed down, or at least half of all checks failed.
• Red means the whole period was down.

The uptime percentage shown on the Monitoring page, a monitor's own page, and your status page is the share of time the monitor was not confirmed down. A check that failed from only one unconfirmed location doesn't lower it — only a confirmed outage does, for as long as it lasted. Planned maintenance never counts against it. On a status page, time during a minor incident counts at 30%, following the usual convention for public status pages; the app and the monthly report count every minute in full.

Filtering a monitor's page to one region changes the percentage to reflect only that region's checks — including unconfirmed failures — which is useful for investigating where a problem happened. The unfiltered figure is the one to quote generally.

The latest checks strip at the top of a check's overview shows recent checks across all locations — hover a block to see when it ran and from which region.

#Read the "Over time" card

Further down a monitor's page, Over time shows everything on one timeline: a row for the monitor overall, a row per location, response times, and events like deploys, changes, incidents, and maintenance.

1. Pick a time range at the top: 24h, 7d, 30d, or 90d.
2. Hover any moment on the timeline to see it reflected across every row at once.
3. Check the percentage at the end of each row — on the Overall row, it's the uptime figure to quote; on a location's row, it's that location's own share of successful checks, which may be lower if it struggled on its own.
4. Look at the two response-time lines: the darker line is the median, the lighter one is the slowest 5%. Red shading marks an incident, grey shading marks maintenance.
5. Check the legend under the timeline to see which colour represents which kind of event. When several events happen close together, the most important is shown first: incident, then deploy, then change, then maintenance.

#Compare regions on the chart Business

1. Open a check's detail page.
2. Open the dropdown above the chart and choose Compare regions.
3. Review the overlaid response-time lines, one per region.

Each line only appears where that region actually reported — a break in a line means no measurement from that region, not missing data overall. This can happen if:

• The probe there was offline or restarting. InfraNest may temporarily borrow a probe from another region to keep your checks and alerts running, crediting the measurement to the region that made it. Hover the gap or read the note under the chart for details (for example, "APAC had no probe of its own for part of this period. It was checked from EU instead").
• The check hasn't been running from that region for very long yet, so it has no earlier history.

#See where the time goes

For website, API, and keyword monitors, you can break down exactly where response time is spent.

1. On a check's detail page, choose Where the time goes above the response-time chart.
2. Review each region's bar, split into: DNS (looking up your domain), Connect (opening the connection — mostly distance), TLS (the secure handshake — also mostly distance), Server thinking (your server preparing the response), and Download (receiving the rest of the page).
3. Read the sentence under the bars that highlights what stands out.
4. If one region is much slower and the extra time is in Connect and TLS, the cause is distance — a CDN or a server closer to that region would help. If the extra time is in Server thinking, your server itself is answering that region more slowly.

TipThe 24 hours view uses every individual check, so it shows brief spikes clearly. The 7 days and 30 days views average results into larger blocks, which smooths out short spikes.

The breakdown is kept for as long as your plan retains monitoring history, and only covers locations that measure each step — others are left out until they do.