Skip to main content

1.7.0

Released 8 Nov 2026

Standalone installations: Prometheus scraping

The bundled nginx no longer publishes /metrics on the public HTTPS port — on a fresh install, and on an upgraded host too: the upgrade removes that one block from the existing site file (your other edits stay). A Prometheus server that scraped https://<host>/metrics from another machine stops receiving data after the upgrade: scrape the backend's loopback port on the host, or set telemetry.metrics.token and add your own proxy location for /metrics. An install whose site file was edited by hand so that the block is no longer recognised keeps publishing it — remove the location = /metrics block yourself.

New​

  • Servers page. Discovery → Servers lists every address and port discovery reached once, however many names it answers for and however many sources scan it, with search, status, source, TLS and key-exchange filters and an export.
  • A page per server. Clicking a server shows its certificate, its TLS settings, every name it answers for with the certificate each one gets, its open findings and the history of everything it served.
  • Tag rules. Discovery → Tag rules tags discovered certificates and sets their owner on their own, from conditions on names, issuer, key type, wildcard, self-signed, public trust, source, address or port. Rules run after scans, hourly and right after a change, show how many certificates they match, and come with one-click examples such as External and Internal.
  • Saved views. The certificate inventory, the Servers page and the Findings page can save the current filters under a name and pin them above the table; each user's views are their own.
  • Activity page. Discovery → Activity shows what changed over the last day, week, month or quarter: servers that started or stopped answering, new certificates, replaced certificates, TLS settings changes, and findings raised, reopened or resolved, grouped by day. A change seen under several names of one server, and findings of one type changed the same day, are shown once.
  • Finding history. Expanding a discovery finding shows when it was opened, resolved by a scan or by hand, reopened, acknowledged or suppressed, its severity changes and who made each change. Changes are recorded from this version on; while older nodes still run in a cluster, their changes are not recorded.
  • Discovery overview. Discovery → Overview shows when discovered certificates expire over the next 13 weeks, how many CertAutoPilot manages with the unmanaged ones that need attention first, the CA, key type and TLS version breakdowns, the week's activity, the most common risk reasons and the certificates served on the most servers. Every number opens the list it counts.
  • Insights. Discovery → Insights groups the discovered estate by issuing CA (with chain trust and CAA), by domain or network, by tag and owner, by risk reason, and by reach — how many servers stop working together if one certificate expires.
  • Scan health of each source. The source list, the source page and the overview show whether a source's last scans were complete, and say what they may have missed, such as names past the SNI limit or a sudden drop in reachable endpoints.
  • Certificate timeline. A discovered certificate's page shows on which servers it was served, from when to when, what replaced it and when TLS settings changed in between.
  • Discovery in ⌘K search. Typing an address, a name or a certificate subject in the search box jumps straight to the discovered server or certificate.
  • Change renewal settings after issuance. Edit on a certificate's Renewal Policy card changes auto-renew, renewal mode, threshold and key rotation, with the same rules as the issue wizard.
  • Audit log filters search the whole log. The action, actor and resource filters run on the server, so an entry on a later page is found; they are kept in the address and the export applies them too.
  • One certificate for names in several DNS zones. A certificate's names can now span many zones — for example eighty provinces' own domains with their www. names — and each name is validated with automatic DNS-01 through its own zone's DNS credential. The issue form groups the pasted names by the zone each one is under, lets you move a name or a group to another zone, and adds www. to apex names in one click.
  • One zone for several domains. A zone can now list many domains served by the same DNS account — eighty provinces' domains under one credential become one zone instead of eighty — with an optional name to show it by, and Edit on the zone list changes its domains, name, credential and description. A domain that a certificate still validates through cannot be removed.
  • Manual DNS records grouped by zone. For a manual-DNS certificate the DNS Challenges card shows the zone each record falls in, lists the records zone by zone, and copies them all as zone-file lines or downloads them as CSV.
  • One tab bar across discovery. Overview, Inventory, Activity, Servers, Findings, Insights, Target candidates, Quantum readiness, Tag rules and Sources are one click apart on every discovery page.
  • Trusting an appliance's management certificate. PAN-OS, F5 BIG-IP, NetScaler and FortiWeb targets can add the appliance's private CA to the verification (ca_cert) or pin the served certificate's SHA-256 fingerprint (tls_fingerprint_sha256), which the health check now shows; targets that set neither work exactly as before. A deploy, dry run or health check that sends a password to an appliance that is neither verified nor pinned now warns TLS_UNVERIFIED_PASSWORD.

Improved​

  • An opened audit entry shows its position in the signed chain, the number to pass as from_seq when verifying the chain from that entry on.

  • PFX files can be made for Windows Server 2016. Certificate downloads, Path Set .pfx files and SMTP PKCS#12 attachments have a new PFX encryption choice: Legacy (TripleDES) opens on Windows Server 2016, the default stays AES-256.

  • Choose the columns of the server lists. Optional columns can be hidden, and the choice is remembered in the browser.

  • A server finding opens its server. Clicking a finding about a server (a missing intermediate, weak ciphers, SNI required…) opens that server's page, and the address under a certificate finding links to its server.

  • Search servers by address and port. The server search finds 10.0.0.5:8443 (and [fd00::1]:443) as the list writes it.

  • Discovery tabs and module guides answer a click at once. The chosen tab or guide is marked straight away with a spinner while the page draws, and Activity shows its entries 100 at a time.

  • The Servers page keeps its filters in the address. A filtered list can be bookmarked or shared, and the browser's back and forward buttons walk through the changes.

  • "Where it can go" on a discovered certificate switches between candidates, added and removed ones with their counts, and has a search and paging; the Target candidates page offers 20, 50 or 100 rows per page.

  • Tag and owner filters, exports and tag suggestions include the tags rules gave. An owner typed on a certificate still wins over a rule's.

  • Older versions of managed certificates are recognised. A server still serving an earlier version of a certificate CertAutoPilot manages shows Older version with a link to the certificate, raises an Older version still served finding and gets discovery expiry alerts.

  • Certificates on servers that stopped answering are marked. The inventory says when a server serving the certificate last answered and can filter them, and the overview tells how many expired certificates are only on such servers.

  • Clearer certificate timeline. A certificate whose servers have not changed since it was first seen gets one sentence instead of identical bars, and a server that stopped answering shows when it last answered.

  • AD CS key size is checked before the request. A template's minimum key size is synced with the template; the issue wizard disables shorter key types and switches to the default key type when the template allows it, otherwise to the smallest allowed one, and a shorter key is refused at once instead of being denied by the CA. Sync the templates again to enable it for an existing connection.

  • Serial numbers in hexadecimal. The certificate page, version history, revoke and rollback dialogs and discovered certificates show the serial the way openssl and browsers do, with the decimal form on hover; the certificate CSV export adds a hexadecimal serial column at the end.

  • The DNS credential test names its record and resolvers. A successful test names the TXT record it wrote, and a zone that cannot be found names the resolvers that were asked.

  • A path set may start with a project variable. ${{ SITE_DIR }}/ssl/fullchain.pem is accepted; the rendered path must be absolute, which the deploy and now also the dry run check.

  • A certificate CertAutoPilot issued is linked by the scan that sees it. A discovered certificate used to stay not managed until a later scan when another source of the project had just scanned.

  • New inventory and server filters. The certificate inventory filters by expiry date range, risk reason, CAA verdict, domain, chain trust, and certificates without a tag or owner; the server list filters by lowest TLS version, network and domain. Filters without a control in the bar are shown as removable chips.

  • Discovery pages share one navigation bar and one header. The areas are grouped with icons, and every page shows its title, a short description and its controls in the same place (page controls, Labels explained, Refresh, then the main action).

Changed behaviour​

  • A user limited to granted resources sees less of the project audit log. Besides its own entries, it now sees only entries about the resources it can see; entries about users, groups, grant changes, the project and its policies made by others are no longer shown, in the list, an opened entry or the export.
  • A stored credential is sent only to the hosts it is used with. Pairing it with a new host, weakening a bound target's certificate check, or changing where a DNS credential, AD CS connection or store-backed credential points is for the person who entered its secret (or set up its secret store) or an organization owner; anyone else enters the secret again first. The credential page shows who entered the secret and the hosts it may reach, and the target form warns before a save that would be refused.
  • F5 BIG-IP, NetScaler and Kubernetes API-server addresses must be plain. The address must be a host name or IP with an optional port, not a URL fragment.
  • Running a distribution that executes commands needs the Admin role. Execute and rollback of a distribution whose SSH or WinRM targets run an action set, and choosing a distribution's path set or action set, are refused for operators (or become an approval request when the project gates their distribution actions). Distributions started automatically after a renewal are not affected, and the Run and Roll back buttons say so instead of failing.
  • The standalone installation no longer publishes /metrics on its public port. Scrape it on the host from the backend's loopback port, or set the new metrics token and proxy it yourself. The metrics token (telemetry.metrics.token) can protect the endpoint in every installation type, and telemetry.metrics.enabled: false now really removes it.
  • The installer and the upgrade remove the old public /metrics location carefully. The block is removed whole, including anything nested inside it, after a backup copy of the nginx site file is taken; if nginx refuses the edited file, the backup is put back unchanged and the log names it so the block can be removed by hand.
  • Changing an appliance target's pinned fingerprint or CA counts as turning its TLS check down. On PAN-OS, F5 BIG-IP, NetScaler and FortiWeb targets with a bound credential, replacing or removing a pin or CA, swapping certificate verification for a pin, or adding a CA beside the system roots is for the person who entered the credential's secret or an organization owner (adding a pin while verification stays on remains open to every editor). On F5 BIG-IP and NetScaler a CA or pin together with use_ssl off is refused.
  • F5 BIG-IP and NetScaler fixed file names are limited to letters, digits, ., _ and - (at most 128 characters). The names reach the appliance's upload address and file commands. A name saved earlier keeps working and is checked only when it is changed.
  • Discovery counts servers, not rows. The inventory's Servers tile, the certificate tables' Servers column, a source's Servers tab and a certificate's Seen line count each address and port once, so the numbers are lower than the endpoint counts shown before.
  • Server rows no longer expand. The inventory's and a source's Endpoints tabs are now Servers tabs, a certificate's Endpoints and TLS card is now Where it's served, and the Quantum readiness list shows servers; the details of each name and the history are on the server's page.
  • The endpoint history drawer is gone. A server's history, across all its names, is on its page.
  • The discovery sources list moved to the Sources tab. Discovery now opens the Overview; bookmarks of the old address land there.
  • Servers that have failed every scan for 30 days count as gone. The next scan removes them with an endpoint gone notification, and a certificate nothing else serves leaves the inventory; the server comes back when it answers again.
  • An older version of a managed certificate is no longer counted as managed or as not managed. Managed now means CertAutoPilot's current version; the older ones have their own count and managed=outdated filter value.
  • The overview's CA breakdown counts self-signed certificates as one part. Each used to count as its own CA.
  • With DNS Resolvers empty, finding the zone uses the public resolvers too. Issuance, renewal, record cleanup and the DNS credential test used to ask the server's own resolver unless a synchronous issuance had run since the start; a server that cannot reach public DNS must list its own resolvers under Settings → General → DNS Resolvers.
  • Renewal settings are checked when changed later. ARI can no longer be set on AD CS or manual-DNS certificates; a create already stored such a request as threshold renewal.
  • The certificate policy accepts only "reject" as its violation action. Another value is refused when the policy is saved; the disabled Require approval option is gone.
  • Issue, Import and editing are shown only to those who may use them. A viewer no longer sees Issue certificate, the edit pencil or the selection boxes, and Import is shown only to admins, who alone may import.
  • ACME certificate names are stored in their wire form. Names are lower-cased, lose a trailing dot and an internationalised name becomes its punycode form (İstanbul.com.tr → xn--istanbul-o0e.com.tr), so such a certificate shows and deploys under that name; two spellings of one name are refused, a * is accepted only as the whole first label, and a name that is not a valid DNS name (an empty or over-long label, a character other than letters, digits, hyphens and underscores) is refused when the certificate is created.
  • Certificate policy rules match a name in either spelling. A domain pattern written in Unicode (*.örnek.com.tr) and a certificate name in punycode match each other, both on the server and in the form's live check.
  • Windows DNS writes and removes challenge records many at a time. A certificate with many names on a Windows DNS zone writes its records in about a minute and a half instead of about thirteen for 100 names, and removes them as fast.
  • Windows DNS gets more time to propagate. When a zone of the certificate uses a Windows DNS credential, issuance on the default (phased) flow waits up to at least two hours for the challenge records to appear (AD replication between domain controllers can be slow), and still continues as soon as they are visible.
  • Manual DNS validation waits for every record. Validate is refused while any TXT record is still not visible, and the message names the missing ones; it used to start as soon as one record was visible, and one missing record then failed the whole order at the CA. Skip propagation pre-check still lets the CA decide.
  • A manual-DNS certificate request with a zone is refused. It used to be stored unchecked although no zone is used; leave zone_id out. A list of zones per name (domain_zones) is refused for manual-DNS and HTTP certificates.
  • Zone domains are checked when saved. A zone's domain must be a valid DNS name without a wildcard, is stored lower-case without a trailing dot (an internationalised domain in its punycode form), and a domain another zone of the project already has is refused.
  • A zone used for some names of a certificate cannot be deleted. The delete check, the zone's certificate list and the DNS credential's usage list count every zone a certificate's names use, not only its first one.
  • A server's TLS version ignores names whose last scan failed. Such a name kept the version of its last success and could make a server look older than it is.
DNS record cleanup jobs can now fail

A cleanup_dns job that finds no DNS credential able to delete the challenge records — the credential was deleted and no zone of the record is known — now ends failed instead of completed, so the leftover _acme-challenge records are visible on the Jobs page. Installations whose cleanup used to pass quietly may see failed jobs after the upgrade; each one names the records that were not removed.

Rolling upgrade

Only 1.7.0 nodes apply tag rules, so during an upgrade certificates can miss their rule tags until a 1.7.0 node runs the next apply.

A certificate whose names use more than one zone can be created only once every node runs 1.7.0 (until then the request answers 409 Cluster upgrade required). Once such certificates exist, do not start an older version again: it would write every name through the first zone's credential and their renewals would fail.

A zone with more than one domain can likewise be saved only once every node runs 1.7.0 — an older node checks names against the zone's first domain only.

Three protections of this release are complete only once every node runs 1.7.0, because an older node does not know them: an operator's request that lands on an older node can still start a distribution that runs an action set, a restricted user reading the audit log through an older node still sees entries about other users' access grants, and a sign-in refreshed by an older node has no session id, so a later sign-out of an API client on a new node does not end that sign-in's refresh token (it ends after the next refresh on a new node, or when it expires).

The first 1.7.0 start builds new database indexes, which can take a while on a large installation. If one project holds the same internationalised zone domain twice (once as written, once in its xn-- form), the start-up log names the second zone; it keeps working, and you should delete one of the two.

Choosing the legacy PKCS#12 encryption on a path set file or an SMTP target answers 409 Cluster upgrade required until every node runs 1.7.0, because an older node would write the modern form.

An approved PKCS#12 download that asked for the legacy encryption is served in the modern form if a node older than 1.7.0 answers it — finish the upgrade before approving such downloads.

Credentials are bound to their hosts once, when the first 1.7.0 server starts; a target saved by an older server after that is refused at its next deploy until someone allowed to adds its host. Finish the upgrade before editing targets.

A PAN-OS, F5 BIG-IP, NetScaler or FortiWeb target saved through an older node loses its ca_cert and tls_fingerprint_sha256 (the older node does not know the fields); the next deploy, dry run or health check then warns TLS_UNVERIFIED_PASSWORD until the target is saved again on a 1.7.0 node.

Fixed​

  • A store-backed credential cannot be created with a ready-made host list by someone other than the store's owner. Only the person who set up the secret store, or an organization owner, may name allowed hosts when creating such a credential; anyone else gets the same refusal as when pairing it with a new target.

  • Re-entering a credential's secret drops the hosts a previous non-owner added, even when the host list is sent along. The removed hosts are reported in the answer, and the new setter can add them back deliberately.

  • Audit verification no longer reports a false break for five minutes after an older entry expires. The retention rule now matches the moment the log removes an entry.

  • The standalone upgrade only removes the public /metrics block. If that block had already been removed by hand and only its comment remained, the upgrade could delete the next nginx location (the HTTP-01 challenge path) instead.

  • A DNS check that runs out of time counts the servers it did not reach as unanswered. Manual-DNS validation then says the resolvers did not answer instead of asking you to create the records first.

  • Switching projects no longer keeps the previous project's Reach insight or zone suggestions on screen.

  • The F5 BIG-IP and NetScaler target forms no longer send a pin or CA when HTTPS is off, and an F5 or NetScaler file name saved before 1.7.0 no longer blocks the edit form.

  • A failed first issuance can be chosen as a notification event. The event was already sent for every issuer, but the rule and template lists did not offer it.

  • Zone domain refusals read plainly. Refusing to remove a zone domain that certificates still use, or to add a domain another zone holds, no longer prefixes the explanation with internal error words ("zone domain in use: status conflict:").

  • A form whose field check failed no longer also shows "Failed to save". Nothing was sent (for example a .pfx row without a passphrase variable), and the field shows its own message.

  • A pinned fingerprint that the appliance does not present is reported as TLS_TRUST_REFUSED and not retried. On PAN-OS, F5 BIG-IP, NetScaler and FortiWeb targets a pin mismatch was reported as the module's connection error and retried; it now fails the target at once with the fingerprint the appliance presented, and the healthy targets of the same run are not rolled back for it.

  • Audit chain verification no longer reports entries numbered late as deleted. An entry written by an older node during a rolling upgrade is numbered after the chain head with an older timestamp; verification now walks the chain by position, so such an entry is checked instead of skipped, and when the 30-day retention removes it from inside the sequence the hole is recognised (every entry records when its predecessor was written) and not reported as a break. ?limit= must now be a positive integer.

  • Deactivating an offline license no longer downgrades an online one. On an installation with an online license the call marked the license invalid and dropped it to the Free limits; it now answers that no offline license is active and changes nothing. After an offline license is removed, a stored online license is validated at once.

  • Signing out ends the session for API clients too. A client that sent its access token as a Bearer header kept a working refresh token after logout, for up to seven days; logout now ends every refresh token of that sign-in, wherever the client keeps it, without touching the user's other sessions.

  • The PAN-OS vsys picker no longer reads the firewall's whole vsys configuration. It read every vsys with its rules and objects; on a large firewall the answer was cut off and part of the configuration (rule names, identifiers) appeared in the picker's error. It now reads only the vsys names, and device configuration is never quoted in an error.

  • Validate on the License page checks the license every time. It used to report success without checking when the automatic check (made when a server starts, at most once a day per installation) had already run that day; it now always checks and shows why when it cannot.

  • A PAN-OS deploy no longer fails as "authentication failed" when the firewall answers "Session timed out" while it is committing. A busy management plane can answer the commit-status poll that way; the deploy now keeps polling until the commit finishes (or the commit timeout passes). A wrong or revoked API key is still reported as an authentication failure.

  • The audit log's tamper-evidence chain is one exact sequence and signs the whole entry. Every audit entry now carries its position in the organization's chain, and its checksum covers the actor, project, resource, values and metadata as well as the identity fields — a rewritten "who did it" or a changed stored value is now a reported break (only the identity fields were signed before). Two entries recorded at the same moment can no longer fork the chain, verification walks it by position, names a reason for every break and pages with next_seq / from_seq; entries from earlier releases are numbered once at start-up and keep verifying under their own formula.

  • Windows DNS works with an internationalized zone. A Windows DNS credential whose zone is typed in Unicode (örnek.com) refused every certificate under it ("not under the configured zone"), because certificate names are in punycode. The two spellings now count as one zone and the server is addressed by the punycode name the zone must be created with; the credential test names a zone that exists on the server under its Unicode spelling only, since no certificate authority can see such a zone's records.

  • Windows DNS batched writes no longer take a record that differs only in letter case for the one being written. When a record failed to add, the read-back compared the value case-insensitively; it now has to match exactly.

  • Windows DNS record cleanup no longer reports success against a missing zone, and no longer fails on a record that is already gone. A zone that did not exist on the server (a wrong zone in the credential, a deleted zone) was read as "record already deleted" and the challenge records stayed behind; a record that was already removed failed the cleanup on every retry. The server's error code now decides, whatever the server's language.

  • IIS and Exchange deployments work on Windows Server 2016. The certificate file sent to the server was AES-encrypted, which Windows Server 2016 cannot open, so every import failed with IIS_IMPORT_FAILED "The specified network password is not correct." It is now encrypted with TripleDES under a random one-time password.

  • The HSTS check and cipher enumeration work on names only your internal DNS knows. They looked the name up with the server's own resolver instead of the Discovery DNS Resolvers the scan used, so on such names the HSTS check never answered and every cipher probe failed.

  • DNS records are removed with the credential that wrote them. When a zone was pointed at another DNS credential while an issuance was waiting, the challenge records were deleted with the new credential and stayed behind.

  • AIA and revocation checks reach an internal CA's web server. The CA host is now looked up with the Discovery DNS Resolvers first and the server's own resolver second.

  • The issue wizard no longer starts on a key type the certificate policy refuses. It selects the organization's default when allowed, otherwise the first allowed key type.

  • A change of DNS Resolvers takes effect at once for finding a record's zone. The zone found for a name was remembered for the zone's SOA refresh time (often hours) whichever resolvers found it, so after switching to internal resolvers on a split-horizon network the challenge record could still be written to, and removed from, the zone the public view showed.

  • A settings read failure during an issuance no longer switches running DNS-01 jobs to the public resolvers. The resolvers in use are kept until the settings can be read again.

  • Check DNS shows a name and its wildcard separately. Both share one record name with two values; the wildcard row showed the other value's state.

  • Manual DNS tells unreachable resolvers apart from missing records. When no DNS server answers, Check DNS shows No answer and Validate says the resolvers could not be reached instead of asking you to create records that may already exist. Copy all on the DNS challenge card also works on a plain-HTTP page.

  • The inventory's server count matches its list. The Servers tab and tile counted servers that failed at the last scan, which the list hides by default (409 on the tab, 151 in the list); they now show the servers answering, and the tile names the total.