Glossary
The words that come up when something breaks. Each one is a definition you can link to, and a pointer to the place on this site that explains it properly.
What things are called
- NXDOMAIN
- The authoritative answer that a name does not exist. Not an error and not a timeout: the resolver was reached, it answered, and the answer was no. More on DNS →
- SERVFAIL
- The resolver could not produce an answer. Something broke between it and the truth, which is a different failure from being told the name is not there. More on DNS →
- DNS propagation
- The wait while resolvers that cached the old record let it expire. Nothing is pushed anywhere; a TTL runs out and the next question gets the new answer. More on DNS →
- DNS hijack
- A record that resolves perfectly, to somebody else. The failure an uptime check cannot see, because the site it reaches is up. More on DNS →
- DNSSEC
- Signatures on DNS answers, so a resolver can tell that what it got is what the zone published. More on DNS →
- DNSBL
- A blocklist you query over DNS, which is why a mail server can check one in the milliseconds it has while your message is still connected. An RBL is the same thing under an older name. More on Email blocklists →
- PTR record
- The reverse lookup: which name an IP address claims. Mail servers care whether it exists and whether it matches. More on Email blocklists →
- STARTTLS
- Encryption arranged inside a connection that began in the clear, which is how mail servers normally do it. Implicit TLS is the other way round: encrypted from the first byte. More on SMTP →
- MX record
- The DNS record naming the hosts that accept mail for a domain, and in what order to try them. More on SMTP →
- Certificate chain
- The certificates between yours and a root the client already trusts. A missing intermediate is the classic version of a site that works in your browser and fails everywhere else. More on SSL/TLS →
- OCSP
- The protocol for asking whether a certificate has been revoked since it was issued, which is a different question from whether it has expired. More on SSL/TLS →
- Certificate transparency
- Public append-only logs of every certificate issued, so a certificate issued for your domain without your knowledge is something anybody can notice. More on SSL/TLS →
- Wildcard and SAN certificates
- One certificate covering every name at a level, or a named list of them. Which names a certificate actually covers is what a hostname mismatch is about. More on SSL/TLS →
- Dead man's switch
- A check that fails when nothing happens. The job calls us, and the alert is the silence rather than the error. More on Heartbeat →
- Pi-hole
- A DNS server people run at home to block adverts and trackers by name. If your domain is on one of the lists it uses, your site does not resolve for everybody behind it. More on Domain blocklists →
- Packet loss
- The share of probes sent that never came back. Partial loss is a real condition in its own right and is not the same as down. More on Ping →
- Timeout
- No answer inside the time allowed, which is a different observation from an answer that took a long time to arrive. More on HTTPS →
- Connection refused
- The host answered the knock and said no. Something is running on that machine; nothing is listening on that port. More on TCP port →
- WebSocket close code
- The number a socket gives when it closes. 1000 is a normal goodbye and 1011 is the server saying it hit an error, and telling them apart is the whole point of watching a socket. More on WebSocket →
What we call things
Words this product uses for its own machinery, which are worth being exact about because they are what an alert and an uptime figure mean.
- Probe
- One of our checking machines. There are three, on three continents and at two hosting providers, and every plan is checked from all of them. Features →
- Confirmation
- The second look. A failure is not an outage until a probe on another continent, at another provider, has seen it too. Most single failures are the network in between rather than your site. Features →
- No data
- We did not observe. A probe we cannot reach tells us nothing about your site, so its silence is never counted as your outage and never alerted on. Questions →
- Incident
- One outage, from when the target first failed to when it answered again, with the evidence that justified calling it one. Features →
- Maintenance window
- Time you scheduled. The check still runs and the result is still recorded; it is left out of the uptime figure and nobody is alerted, because time you chose is not time you suffered. Questions →
- Bearer token
- The credential the public API takes, in an Authorization header. You create it, we show it once, and we store only a fingerprint of it. API →