Monitoring behavior
Check intervals
| Plan | TLS endpoint interval | Domain interval |
|---|---|---|
| Free | Hourly | Daily |
| Starter | Hourly | Daily |
| Pro | Hourly | Daily |
Every plan checks TLS endpoints hourly. Domain checks run on one daily cadence for every account regardless of plan — registration state doesn't change fast enough to justify hourly polling the way a certificate's remaining lifetime can. Adding or resuming an endpoint or domain also triggers an immediate check in the background, so a new entry doesn't sit with an empty status until the next scheduled run. An endpoint's hostname and port can't be changed: to watch another host, add it and remove the old one.
Each endpoint check connects to every address the hostname resolves to, IPv4 and IPv6, up to 8, so your servers see one TLS handshake per address per check. The status comes from the first address that serves a certificate, so a check fails only when no address serves one.
Statuses and reasons
An endpoint or domain has one status, and every status except Healthy shows a Reason beside it, unless the status comes from the renewal or registration expiry alone, which the expiry date beside it already says. The status says what to do; the reason sums up the worst finding behind it in a few words, such as "Certificate problem" for a weak key. Hover over a status, or tap it on a phone, for what it means.
| Status | Meaning | Example reasons |
|---|---|---|
| Healthy | No certificate or registration problem was found: the certificate is valid and its renewal is not due yet, or the registration is not close to expiry. A certificate that renews on schedule is Healthy for its whole life. | None. |
| Late | Past the renewal point and not renewed, with nothing broken yet: a certificate past the point its renewal was expected (see when a certificate is Late or Urgent), or an address still serving an older certificate past that point. A certificate renewed automatically on schedule never reaches it. For a domain: 30 days or less left and not renewed; nothing is broken yet. | "Certificate problem", or none for an overdue renewal or a registration close to expiry |
| Urgent | For a certificate, someone acts today: not renewed and expiry is close, or it fails some clients now or soon (a missing intermediate, an intermediate that expires first, a weak key, a SHA-1 signature, a lifetime over 398 days, or a CA that wants it replaced early). For a domain: 7 days or less left. | "Certificate problem", or none for a certificate not renewed with expiry close or a registration 7 days from expiry |
| Down | Clients reject it now: an expired certificate or registration, or a certificate that does not match the hostname, is self-signed, is not valid yet, or does not chain to a trusted root. | "Certificate expired", "Certificate chain is invalid" |
| Unreachable | The latest checks were not conclusive, so CertWatch cannot vouch for the certificate or registration. The Reason says why. | "Host unreachable", "CertWatch could not finish the check", "Registry lookup failed", "Registry response had no expiration date", "Checking…" |
An Unreachable row reads "reason since date; last known status age", for example "Host unreachable since Oct 3; last known Down 2 days ago". A monitor that is waiting for its first check is Unreachable with the reason "Checking…". If that check misses its deadline, the monitor is Unreachable with the reason "First check overdue since" the deadline, and its last known status if it has one. If CertWatch's own scheduled checks are delayed, a monitor keeps the status its last check gave, and the Dashboard counts it as that status, moved up when its certificate's or registration's Late, Urgent or expiry date passes (an Unreachable monitor stays Unreachable until its Urgent date), and its row shows that status with its reason; nothing alerts until the checks resume. A monitor with no check since it was added or resumed is Unreachable with the reason "First check overdue" instead. A monitor whose checks fail keeps its last status until it is Unreachable (for an endpoint, when the follow-up check about two minutes after a failed check fails too; for a domain, after 24 hours), and its row says so: "Host unreachable since Oct 3". A check CertWatch could not finish on its own side ("CertWatch could not finish the check") never counts toward Unreachable: the endpoint keeps its last status, moved up by its certificate's dates. An endpoint CertWatch has never read is Unreachable from its first failed check. On a monitor's row two states carry their own badge, Pending or Unsupported, in place of the plain status badge. A domain on a TLD for which CertWatch has no RDAP source that publishes an expiry date or on an internationalized (IDN) TLD is Unreachable with a reason that says so: a coverage gap rather than an outage. Supported TLDs lists which TLDs are covered. A paused monitor has no status: it shows Paused, and is left out of the status counts and the Triggered list. Pausing ends its open problem without a Recovered, while its checks and History are kept for your plan's retention. Resuming starts it again from no status: it waits for its first check, which runs at once.
Findings
A finding is something a check found wrong with a certificate or a registration, such as an overdue renewal, a missing intermediate, or a name the certificate does not cover. One check can have several. The status is the worst of them and the Reason sums it up; the monitor card's Overview tab lists every finding of the last check in its "Why it's" box (such as "Why it's Late"), and an alert words each one. A finding on only some of a hostname's addresses names them.
A finding that comes from what a check saw, rather than from the clock (a chain that does not verify, a weak key, a name mismatch, an older certificate still served on some addresses), counts only once it is confirmed: two checks in a row see it. CertWatch usually checks again a couple of minutes after the check that first saw it. Until then the panel lists it as awaiting confirmation, and it changes neither the status nor any alert, so one bad backend answering once does not page you. A finding confirmed by the check after one you started (such as Check now) sends its alert as soon as it is confirmed. An overdue renewal and an expired certificate need no confirmation.
Under "Also noted" the panel lists facts that never change the status or send an alert: a renewal and what changed with it (issuer, a new key, names), an older certificate that is still on schedule on some addresses, addresses that could not be read while another served the certificate, a day count that does not fit the certificate, and a CA renewal window that has passed with no renewal seen.
When a certificate is Late or Urgent
A certificate's Late and Urgent points are measured from when this endpoint's renewal was expected, not from a fixed number of days, so a certificate that renews on schedule never reaches either one. Late means the certificate is past its renewal point and not renewed; Urgent means it is still not renewed and its expiry is close. The expected renewal comes from the best evidence the endpoint has, in this order:
- Your renewal mode, when you have said the certificate is renewed by hand or renews with a set number of days left (see Renewal mode).
- The CA's suggested renewal window (ARI), once CertWatch has seen this endpoint's renewal client renew inside it.
- The renewals CertWatch has seen on this endpoint. Before it has seen one, the hostname's earlier certificates in public Certificate Transparency logs stand in; CertWatch looks those up soon after the endpoint is added, and if the search does not answer, tries again within the hour. A single renewal that came within a week of expiry (or a twentieth of a short certificate's lifetime) is not learned on its own, so one late renewal does not move Late to hours before expiry for good; the endpoint keeps the default, and the API offers that renewal's day count as a setting. A replacement made while the old certificate still had more than half its lifetime left, such as a reissue after the server was rebuilt, is not learned from.
- A default, while none of the above is known. The renewal mode on the monitor card's Settings tab then says CertWatch is learning, and the dates tighten after the first renewal it can learn from.
In automatic mode, Late is never earlier than the start of the CA's suggested renewal window for the current certificate, once CertWatch has fetched one; when that start sets the date, the card and the alert name the date "the CA's recommended renewal start".
Once CertWatch has read the certificate, the monitor card's Overview tab shows both dates, as Late from and Urgent from (the labels alerts use), which evidence they come from, and the CA's renewal window when there is one, while the endpoint is active. If the dates do not match how the certificate is really renewed, set the renewal mode on its Settings tab, before any alert goes out. Renewed by hand is Late 30 days before expiry (half the lifetime on a certificate of 60 days or less), where the card's finding reads "Time to renew", and "renews with N days left" is Late one day after N. Urgent is 7 days before expiry, or halfway from Late to expiry when that is closer to expiry. A day count as long as the certificate's lifetime or longer is refused when you save it; if a later certificate is shorter than the day count, it is set aside for automatic, and the card says so.
The findings in the table above make a certificate Urgent or Down whatever its dates say, and an expired certificate is Down. A certificate with no known start date has no renewal dates, and is still Down once it expires. Endpoints are checked hourly on every plan, often enough for a certificate that lives only a few days.
Renewal mode
The renewal mode says how an endpoint's certificate is renewed. Set it on the monitor card's Settings tab, where each mode shows its Late and Urgent dates before you save:
- Automatic: dates follow the renewals CertWatch sees, or the CA's suggested window. Every new endpoint starts here.
- Renewed by hand, for a certificate a person orders and installs: Late 30 days before expiry, or halfway for a shorter certificate, so a person has time to act (see when a certificate is Late or Urgent).
- Set days left, for automation that renews with a set number of days left: Late once fewer days are left than you set.
Saving a new mode or day count checks an active endpoint again with the new setting ("Checking now."), and its dates and status follow that check within seconds; a paused endpoint has no dates until you resume it. A new status that check finds sends its alert at once; reminders follow the scheduled checks. The monitor card's Overview tab names the mode under the renewal points, and the CA's suggested window ("The CA suggests renewing Nov 16–18.") only when the dates come from it. Each choice is recorded on the endpoint's History.
When a domain registration is Late or Urgent
A domain registration is not renewed on a cycle, so it keeps fixed points, each an exact moment before its expiry: Late from 30 days before it, Urgent from 7 days before it, and Down from the moment it expires. The first check past a point shows the new status.
A new certificate
A new certificate, issuer, key, or set of names is what a healthy renewal looks like, so it never sends an alert. It is a "New certificate observed" entry on the monitor's History, and the Renewal panel notes the renewal and what changed with it. A new certificate that breaks something (a missing intermediate, a name it no longer covers) is caught as a finding and alerts by its status.
Triggered
A monitor is Triggered when its status is Late, Urgent, Down, or Unreachable. It is not Triggered when it is paused, still within the time allowed for its first check, or a domain on a TLD with no registration data. A monitor whose first check is overdue is Triggered. A monitor whose scheduled checks are delayed is Triggered only when its status is. The Dashboard lists the Triggered monitors (see Dashboard). Triggered is worked out from the status each time it is read, is never stored, and does not send anything: alerts follow their own rules (see Alerts).
Check history and retention
| Plan | Check results retained |
|---|---|
| Free | 30 days |
| Starter | 90 days |
| Pro | 365 days |
Every check is recorded, not just the latest result, so you can see when a certificate was last reissued or when a domain's expiry moved. Check results older than your plan's retention window are pruned automatically based on your current plan — so downgrading from Pro to Free will retroactively trim stored checks to match the Free plan's 30-day window on the next prune cycle. The History of changes to a monitor is separate and keeps 90 days on every plan. CertWatch also keeps a short record of each certificate a monitor has served (its serial, fingerprint, and when it was first and last seen), up to 20 per monitor and outside the retention window, to learn when the monitor's certificates usually renew. It also keeps every renewal CertWatch finds (when it happened, the time left, the issuer, and the key algorithm), for the life of the monitor on every plan, outside the retention window, up to the last 500 per monitor. Removing the monitor removes them.
Muting
An endpoint's Mute menu, in its monitor card's header, on its Monitors row, and in the Dashboard's Triggered pane, mutes it for 1 hour, 24 hours, 7 days, 30 days, or Until I unmute, and Unmute ends a mute early. A choice that would still hold when the certificate expires says "Ends after the certificate expires on" that date, since the mute would hold the Down alert; Until I unmute always does once CertWatch has read the certificate. Muting holds every alert, and holds the Recovered for a problem that ends during the mute, which is not sent when the mute ends; checks, status, and history keep updating regardless, so a muted endpoint's dashboard tile is never stale, only quiet. Muting is separate from Acknowledge, which holds back an open alert's reminders; that section says what ends it.