Targets & groups
A target is one concrete endpoint of a given module type — an SSH host, a K8s cluster, an F5 BIG-IP management URL, a Vault instance. A target group is a named collection of targets for bulk distribution. Both live under Settings → Distribution.
Creating, editing, and deleting targets and target groups (including group member changes) is operator-capable, gated by default through the configuration changes rule of the project's approval policy (Settings → Approvals, all operators by default) — an operator's change is filed as an approval request an admin approves before it applies. Admins and owners bypass approval. Module credentials referenced by a target stay admin-only. See Approval workflow and Auth & RBAC.
Without a licence the installation can hold one target at a time, counted across all projects. Creating a second one is refused (403 target_limit_reached); deleting the target frees the slot. See License.
Model
- Identity vs display: every target has an immutable
_id(used in jobs, audit, references) and a mutable_name(safe to rename for humans). References use IDs; UI surfaces names. - Module type: fixed at creation time. Change requires deleting + recreating the target.
- Credential reference: most modules authenticate via a module credential. Multiple targets can share one credential (e.g. ten SSH hosts, one key). Rotating the credential rotates it for every target that references it.
- Project scope: targets belong to a single project. Cross-project reuse is not supported — create a target per project.
Create a target
The Targets list can be searched by name and narrowed with the type filter next to the search box (both apply server-side across every page).
A target can also start from what discovery found: Discovery → Target candidates → Add as target opens this form with the address and validation endpoints filled in — see Target candidates.
- Settings → Distribution → Targets → New.
- Pick a module type. The form reveals type-specific fields:
- SSH — hostname, port, username, credential reference (SSH key or password), and an optional host-key verification toggle.
- Kubernetes — kubeconfig credential or in-cluster mode, default namespace, explicit TLS Secrets (empty = one auto-named Secret per certificate) and restart targets — or, instead of listing them,
placement.mode: match(Find by certificate names), which finds the TLS Secrets the certificate covers and rolls out the workloads that use them, see Find the Secrets by certificate names; namespaces, Secrets and workloads can be picked from the cluster. - IIS — WinRM host + port, auth type on the target (NTLM / Basic / Kerberos with realm + optional KDC pin; empty = the credential's legacy auth fields), credential reference (bare username + password), certificate store + binding placement — site, bindings and application pools can be picked from the server, with a preview of the bindings a deploy would touch.
- WinRM (generic Windows) — hostname + port (empty = 5985 plain / 5986 TLS), TLS toggle + skip-verify, auth type (NTLM / Basic / Kerberos) with NTLM domain or Kerberos realm (Kerberos is zero-config — optional KDC pin, no krb5.conf), credential reference, PathSet (where files land) and ActionSet (post-deploy script), connect/command timeouts, max file size. The Windows sibling of the SSH target.
- F5 BIG-IP — management host + port, partition, credential; optional
ca_cert(private CA, withssl_verifyon) ortls_fingerprint_sha256(pin of the served certificate) — see F5 BIG-IP. - NetScaler (Citrix ADC) — NSIP, credential, intermediate naming strategy (fingerprint-based recommended; see NetScaler); optional
ca_cert/tls_fingerprint_sha256as for F5 — see NetScaler. - HashiCorp Vault — Vault address, KV mount path + engine type (
kv2orkv1), secret path, credential (token or AppRole). - Webhook — URL, HTTP method, headers, timeout, retry policy.
- Huawei Cloud — region, project ID, credential (AK/SK), scope (ELB / CDN / WAF).
- AWS ACM — region (CloudFront requires
us-east-1), credential (access key), optionalcertificate_arnto re-import an existing ACM certificate in place — pickable from the region, which also flags the certificates ACM issued itself and therefore cannot be imported over, optionalcertificate_nameused as theNametag when a new certificate is created. - Email/SMTP — relay host + port, TLS mode (
starttls/smtps/plaintext) + skip-verify, auth method (plain/login/none) + username, credential reference, from address + display name, TO/CC/BCC recipients, payload mode (pem/pem_and_key/pkcs12), subject prefix, timeout.pkcs12additionally requires a PKCS#12 passphrase credential (typesmtp_pkcs12_passphrase), and takes an optionalpfx_encryption:modern(default, AES-256) orlegacy(TripleDES-SHA1, for older Windows that cannot open an AES PKCS#12). - Cloudflare — target zone ID (pickable from the account, with the zone's plan shown because a custom certificate needs Business or Enterprise), credential (API token),
bundle_method/type/geo_restrictions, and what to do when another certificate already serves the names — take one over (certificate_id) or upload alongside it (allow_overlap). Uploads the certificate to the zone as a Custom Certificate (Business/Enterprise plan); updates in place on renewal. - cPanel / WHM — host,
mode(cpanel/whm), optional port/use_ssl/ssl_verify/timeout,domains(optional — a distribution can supply them per certificate),whm_user(deprecated, never sent); credential (API token{username, api_token}). Installs the certificate onto each domain via cPanel UAPI / WHM API; overridable domain list per distribution. - Azure Key Vault — vault URL (
https://<name>.vault.azure.net; sovereign clouds supported), credential (service principal), optionalcert_name(empty = per-certificate auto name), optional tags + HTTP timeout. Imports an AES-256-encrypted PKCS#12 (random per-import password) as a new version of the KV certificate on every renewal. - MerlinCDN — binding rows (
bindings: distribution ID + canonical-name ID or domain per row; one upload, each CNAME re-pointed), credential (personal access token + org/workspace), optional API host,bindtoggle (off = upload-only),delete_old_certtoggle. - Microsoft Exchange — WinRM host/port (address the server by its AD FQDN), Kerberos-only auth with the
domain(realm) required and an optionalkdc_hostpin, services (IIS/SMTP/POP/IMAP), apply mode (single / explicit list / all org),restart_services+delete_old_certtoggles, credential (bare username/password). The server list can be picked from the organization. With restarts on, every server in the list is contacted directly, so name each one by its FQDN and allow WinRM to all of them. Exchange serves RSA certificates only. On-prem Exchange Server only. - PAN-OS (Palo Alto) — management host + port, credential (
username+passwordorusername+api_key; username always required), optionalcert_name(empty = per-certificate auto name), optionalvsys(empty = shared location) — both pickable from the firewall,auto_committoggle (off = import to the candidate config only; the run succeeds with a "commit pending" warning),ssl_verifywith optionalca_cert(private CA) ortls_fingerprint_sha256(pin of the served certificate — see PAN-OS), timeouts. Imports cert + key + chain as one named certificate object, then runs a partial commit scoped to the API user; renewals re-import under the same name, so SSL/TLS profile and GlobalProtect bindings are preserved. Requires PAN-OS 9.0+. - FortiWeb (Fortinet) — management host + port, credential (
username+password; no API token exists), optionalcert_name(updated in place on FortiWeb 8.0.7+,<name>-<sha8>per version on 8.0.5–8.0.6), optionalchain_group, optional placements —snigroup members andmulti_certificatesentries — all pickable from the appliance,delete_old_certtoggle,ssl_verifywith optionalca_cert/tls_fingerprint_sha256(see FortiWeb), timeout. By default no change drops a connection and server policies are never written;update_policies(off by default) allows the policy writes that do. Optionaladom— the administrative domain to work in when ADOMs are enabled (empty = root). - A10 Thunder — management host + port, credential (
a10_password: username and password), optionalpartition(empty = shared), optionalcert_name(max 63, empty = per-certificate auto name), optional client-ssltemplatesandsnientries — pickable from the device, with a deploy preview —delete_old_certandsave_configtoggles,ssl_verifywith optionalca_cert/tls_fingerprint_sha256, timeout. Certificate, key and chain are one file replaced in place on every renewal, so templates and virtual ports keep serving it. See A10 Thunder. - FortiGate (Fortinet) — management host + port, credential (REST API administrator token), optional
vdom(empty = root), optionalcert_name(max 35, empty = per-certificate auto name), optional placements — VIPs, SSL inspection profiles (replace mode only) and the SSL VPN server certificate — pickable from the firewall,delete_old_certtoggle,ssl_verifywith optionalca_cert/tls_fingerprint_sha256, timeout. Creates the local certificate once and updates the same object in place on every renewal, so its references stay; placements are only added, never removed. See FortiGate. - Elchi — controller URL (https only), credential (
username+passwordof a dedicated service account), optional project, optionalsecret_name(empty =cap-<token>-<slug>). CertAutoPilot keeps its own secret in every version Elchi serves (no version is typed) and updates only those. See Elchi.
- Give the target a descriptive name.
prod-web-lb-1beatstarget-42. - Optional validation endpoints (post-distribution TLS check):
- Host + port (
443for HTTPS). - SNI (optional; defaults to the endpoint's own hostname — empty only when the endpoint is IP-only).
- Method:
tls_fingerprint(compare the served leaf against the certificate just deployed) ornone(skip). There is no fingerprint-source choice — the expected value is always the certificate this run deployed. - Retry policy: count, delay, timeout.Validation endpoints can also be defined per-distribution later; per-target is the common case.
- Host + port (
- Save. Saving does not run a health check — it is a separate action on the target (project admin). Run it to confirm the endpoint is reachable and the credential authenticates; failure shows a red badge with the reason.
Add targets in bulk
Settings → Distribution → Targets → Add in bulk creates many SSH, WinRM or IIS targets from one list of servers. Other target types are added one at a time.
The form is the ordinary target form used as a template, with the name and the address replaced by a list:
-
Pick the target type.
-
Paste the servers, one per line. A line holds only what each target must have of its own and a distribution override cannot change later — the name and the address, in that order:
Line Meaning 10.10.4.23address only; the target is named after it — the first label of a host name ( db-ank-01fordb-ank-01.corp.local), the whole address for an IPweb-ist-01 10.10.4.21name, then address mail relay, 10.10.9.4a comma (or a semicolon or a tab) when the name has spaces; rows pasted from a spreadsheet work as they are A line never carries a port, an IP beside a host name, a site or any other setting: those are shared. Up to 500 servers per request.
-
Under the list, a table says what each line would do before anything is created: New, Skipped (the same address appears on an earlier line, or it is already a target of this type — the existing target is named), or why the line cannot be used (not an address, a port in the address, a name already in use).
-
Fill in the settings every server shares — credential, port, sign-in, PathSet and ActionSet, the IIS placement, tags — exactly as on a single target. Live pickers read the first new server of the list. Validation endpoints are per server and are not offered here; add them on a target, or in a distribution override.
-
Optionally choose a target group of the same type; the new targets join it, and so do the servers of the list that are targets already (the table says which of them join it and which are in it already). Adding members is an edit of the group, so it needs the same right as Add Targets on the group (operator on the group); without it the request is refused and nothing is created.
-
Create N targets (with a group, the button also counts the existing targets that join it; a list with nothing new can still be sent when one of its existing servers is not in the group). The result lists every line: created, skipped, or refused with the server's reason. Lines are independent — one refused server does not stop the others — and Correct the lines and try again puts only the refused ones back in the list.
Good to know:
- Running the same list again creates only what is missing, because addresses that are targets already are skipped. With a group it also puts every server of the list into the group — the way to finish a run whose targets were created but not added to the group. Only targets the caller can see count as existing. A target of another type on the same address does not count.
- Nothing is deployed and no server is contacted — as with a single target, run a health check afterwards.
- Approval. An operator who needs approval for configuration changes raises one request for the whole list; the reviewer sees the template and the servers. A template that can make no target, or a group the requester may not add to, is refused at once — no request is raised. If some servers are refused when the approved request runs — also one whose name was taken by another target in the meantime — the request ends as failed although the others were created; executing it again tries only the rest. A caller who is an admin for creating targets but only an operator of the chosen group (a role limited to granted resources) needs approval too, as for adding targets to that group.
- Licence limit. On a plan with a target limit, creation stops at the limit and the remaining lines are reported as not attempted.
- Credential hosts. Each address is checked against the credential's allowed hosts like any target; a server the credential may not be sent to is refused on its own line.
- The request itself is refused (400, nothing created) for another target type, an empty list, more than 500 servers, a missing template, validation endpoints in the template, or a group that does not exist or holds another type.
Health check
Clicking Health check on the target detail page re-runs a minimal reachability test:
- SSH: connect, then open and close an SFTP session (no command is run on the host).
- K8s:
GET /versionon the API server. - F5: iControl REST version query. NetScaler:
GET /nitro/v1/config/nsversion. - IIS: WinRM connect + a preflight script that checks what a deploy needs (Full Language mode, PowerShell 3+, an administrator token, WebAdministration, W3SVC, the machine certificate store, a writable
C:\Windows\Temp); non-blocking findings are reported as warnings. See Health checks. - WinRM (generic): WinRM connect + the same preflight (Full Language mode, PowerShell 3+ fail the check; a non-administrator account, an unwritable
%TEMP%or a lowMaxEnvelopeSizekbare warnings). - Vault: authenticate with the credential, then probe the mount's tune endpoint.
- Webhook: HTTP
HEADto the base URL (reachability probe; status code not evaluated). - Huawei: cert-service list call.
- AWS ACM:
acm:ListCertificatesin the target region. - SMTP: dial the relay, negotiate TLS as configured,
EHLO+QUIT— deliberately no AUTH attempt, so probes don't fill the relay's audit log with failed logins. - Cloudflare: reads the target's zone with the API token (
GET /zones/{id}), which proves the token is valid and reaches that zone. It does not useGET /user/tokens/verify— that route rejects account-owned tokens even though they deploy fine. - Azure Key Vault: AAD token acquisition + certificate list probe (
GET /certificates?maxresults=1) + one per-object read (GET /certificates/{name}, where404is the healthy answer). - cPanel / WHM: authenticated probe (cPanel installed-hosts list / WHM version).
- MerlinCDN: authenticated distribution lookup.
- Exchange: WinRM connect + Exchange Management Shell availability probe.
- PAN-OS: authenticate + read system info (reports the PAN-OS version); fails if the target sets a
vsyswhile multi-vsys is disabled. - FortiWeb: reads the global settings in the target's ADOM (reports
hostname:<name>andadom:<name>, oradom:disabled), the HA mode (reported, not judged) and the firmware version (which update method applies; before 8.0.7 a warning; unreadable ⇒ unhealthy, since a deploy stops on it too). - A10 Thunder: logs in, selects the partition, reads the version, the certificate list and the HA state (reports
acos:,serial:,partition:andha:;A10_HA_UNKNOWNwhen aVCS / VRRP-A is configured). - FortiGate: reads the system status and the certificate list (reports
version:,serial:andvdom:); unhealthy when the target names a VDOM while VDOMs are disabled. - Elchi: logs in (no one-time code), finds the versions Elchi serves (with an Editor account, the project's) and reads the project's secrets.
For PAN-OS, F5 BIG-IP, NetScaler, FortiWeb, FortiGate and A10 a successful check over TLS also lists tls_fingerprint_sha256:AB:CD:… — the served certificate's fingerprint, the value to paste into the target's tls_fingerprint_sha256 pin — and warns TLS_UNVERIFIED_PASSWORD when the target sends a password without verifying or pinning the appliance's certificate (FortiGate: its API token, a reusable administrator credential, is warned the same way).
Health-check failures do not disable the target — distribution still attempts it — but the list view shows a red dot so you have a warning before execution.
Create a target group
- Settings → Distribution → Target groups → New.
- Name it (e.g.
prod-web-lbs). - All targets in a group must share the same module type — you cannot mix SSH and Kubernetes in one group.
- Pick members from the list. The list offers only targets of the chosen module type and searches on the server as you type; switching the type clears the picked members and the default credential. Member count is shown in the list view.
- Optionally set a default credential (the picker lists only credential types the group's module accepts). A member target whose own credential is blank uses it. It is validated like a target's credential: it must exist, belong to this project, and be a type the group's module accepts — a Vault token on an SSH group, or a credential from another project, is refused (400) on create and on update.
On an attached distribution, you pick a single target or a single group, not both. Distributing to a group of 50 triggers fan-out only if a positive fan-out threshold is configured and 50 ≥ that threshold (default 0 = fan-out disabled; Settings → General).
Edit / delete
- Rename is always safe — it updates the display name; the immutable ID stays.
- Add targets to a group (group list → Add Targets) — the dialog lists only targets of the group's module type and has a search box, so a target beyond the first page can be found by name.
- Change credential — allowed; all future distributions pick up the new credential on next execution.
- Change module type — not allowed. Create a new target and update the distribution binding.
- Delete — blocked while the target is a member of a target group, or is listed in a module configuration. A distribution references a module config, never a target directly, so remove it from the group / module config first.
Variables in target config
Target connection fields (hostname, host, IP, NSIP, etc.) are literal values — they are not templated or variable-substituted. Project variables (Project variables) are resolved only in SSH/WinRM PathSet paths and ActionSet commands/scripts, via ${{ NAME }} placeholder substitution (not Go templates) — see primitives. For staging vs prod where only the hostname differs, create a separate target per environment.
Security notes
- Targets are scoped to a project; no cross-project leakage.
- Credential references are by ID — revoking a credential (admin-only) does not delete the target, but the next execution will fail with
authclassification. - SSRF: target hostnames are resolved via the standard network policy — link-local and cloud-metadata blocked by default.
Troubleshooting
Health check is red but I can reach the target manually
Usually credential scope or network path. Check the exact error on the detail page. Common causes:
- K8s: API server unreachable from the backend, or an expired/invalid kubeconfig. (A missing
secrets: create, updatepermission does not turn the health check red — it exercises no RBAC — and only fails at deploy time.) - SSH: wrong port (22 vs 2222), wrong credential, or firewall between the backend pod and the host.
- IIS: WinRM not configured (
Enable-PSRemoting -Forceon the Windows side) or firewall on TCP 5985/5986.
Cannot delete target
The target is still in a target group or listed in a module configuration. Go to Settings → Distribution → Target groups / Modules, remove it there, then retry.
"module type mismatch" when adding a member to a group
Groups are homogeneous. Create a second group for the other module type. If you want to distribute to both types for the same cert, attach two distributions to the cert (one per module type).
See also
Video tutorials
Short, hands-on videos showing this feature working for real. English audio, English and Turkish subtitles. All video tutorials →