Sage Infrastructure Solutions Group IncSage Infrastructure Solutions Group IncLoading…

Sage Infrastructure Solutions Group Inc

Services

A service is one application behind NullWard. It has:

  • one or more hostnames that route to it,
  • one or more destinations (backends) that requests are sent to,
  • the security and access settings that apply to its traffic.

Requests are routed by hostname. Each hostname belongs to exactly one service.

Add a service

  1. Go to Services > Services and click Add Service.
  2. Fill in the form:
    • Name: a label for the dashboard.
    • Host Pattern: the main hostname, either exact (app.example.com) or a wildcard (*.example.com).
    • Additional hostnames: any other names that should reach the same application.
    • Upstream Type: Direct HTTP/HTTPS for a backend NullWard can reach, Application Tunnel for a backend behind a tunnel, or TLS passthrough for a TLS service NullWard must not decrypt, such as an SSL VPN gateway (see TLS passthrough).
    • Upstream URL: the backend address, for example http://10.0.1.20:8080 or https://app-backend:8443.
    • Strip Path (optional): a path prefix to remove before forwarding, for example /api.
  3. Choose the toggles. WAF Enabled and TLS Enabled are off by default, so turn them on for internet-facing applications. WAF Enabled turns on the WAF in Block for both requests and responses; you can set each direction on its own afterwards.
  4. Click Create Service.

The Add Service dialog

Then open the service (click its name, or the edit icon) to configure the rest.

The services list

The service settings page

A service's settings are grouped into tabs (NullWard 0.6.105 or later):

TabSections
GeneralGeneral (hostnames, destinations, load balancing and health, path routes), Maintenance
WAFWeb Application Firewall, WAF Exclusions
AccessAuthentication, IP Allowlist, Path Access Rules, Sessions
Threat protectionBot Detection, IP Intelligence, ML Security, Malware Scanning, Rate Limiting, Exception Auto-Approval, Behavior: HTTP Error Scoring
TLSTLS Configuration
Traffic & limitsQueue / Load Management, gRPC, Compression, Static Asset Cache, Request Audit Logging, Request Size Limits, Proxy & Timeouts, Upstream Forwarding Headers, Upstream Cookie Security

A TLS passthrough service shows only General, Access and Threat protection. The open tab is part of the address (?tab=waf), so you can bookmark or share it.

  • Search settings (top right) finds a setting in any tab by its name, help text, tooltip or a related word (for example country, timeout or OIDC). It shows the matching sections and highlights the settings; Esc clears it.
  • The Access and Threat protection tabs explain in what order a request is checked, with each check's state for this service. Click a check to open its settings.
  • Lists such as bypass paths and exempt paths take several entries at once (NullWard 0.6.107 or later): type or paste them one per line and press Add (Enter adds them, Shift+Enter starts a new line). Invalid regular expressions stay in the box with the reason, duplicates are skipped, and Edit as text edits the whole list. IP and CIDR fields take one entry per line or commas.
  • Save Changes in the bar at the bottom saves every tab at once; there's no restart. The bar shows Unsaved changes until you save, and leaving the page asks first. If a setting is invalid, saving opens its tab and marks the tab with a red dot.

The General section of a service

General: hostnames and destinations

Destinations lists the backends. Each destination has:

  • Type: Direct HTTP or Application Tunnel.
  • URL: for direct destinations, the backend address. For tunnel destinations, pick the Tunnel and Upstream Group instead.
  • TLS Verification: Inherit service, Skip verification (for self-signed backend certificates) or Require valid cert.
  • Protocol (HTTPS destinations): HTTP/1.1–2 or HTTP/3. See HTTP/3.

Destinations can be turned off individually and reordered. The order matters for failover.

Load balancing and health

With two or more destinations, Load Balancing & Health offers:

  • Policy: Round-robin, or Failover (always use the first healthy destination in list order).
  • Sticky Sessions: off by default. When on, each visitor stays on the same backend: the same destination (cookie nw_upstream), and, through an application tunnel, the same pod or target behind the connector (cookie nw_affinity). Turn it on for applications that keep sessions in a backend's memory or issue a session cookie per backend, such as the Wazuh dashboard. Your application never sees either cookie. Tunnel affinity needs connectors 0.6.85 or later. The toggle also appears for a service whose only destination is a tunnel.
  • Passive Cooldown (s): after a connection to a destination fails, the destination is skipped for this many seconds. This works even without an active health check.
  • Active Health Check: probes a Path every Interval seconds. A destination is marked healthy or unhealthy after the configured number of passes or failures. Responses matching Expected Status (default 200-399) count as healthy.

Path routes

Path Routes (in the General section) send requests under a path prefix to other destinations, for example /nats to a different tunnel or backend. Each route has:

  • Path Prefix: for example /nats. It matches that path and everything below it, on whole path segments: /nats and /nats/ws match, /natsx doesn't.
  • Destinations: direct URLs, tunnels or both, like the service's own, with their own Policy (Round-robin or Failover).
  • Strip prefix: removes the prefix before the request is forwarded, so /nats/ws reaches the destination as /ws. The service's own Strip Path doesn't apply to routed requests.
  • Enabled: turn a route off without deleting it.

The longest enabled prefix wins; requests that match no route use the service's destinations. WAF, authentication, rate limits and the service's other settings apply to every route. Route destinations use passive health only (no active health check). A service can have up to 64 path routes. Path routes need NullWard 0.6.86 or later.

Maintenance

  • Maintenance Mode shows visitors a maintenance page, with an optional message and expected end time.
  • Scheduled Maintenance Windows turn it on automatically, once or on a recurring schedule. Each window has a mode: Full Block (all traffic), or Intercept Errors (5xx only), which shows the maintenance page only when the backend returns server errors.

WAF and authentication

  • The WAF tab sets WAF: Requests and WAF: Responses (Block, Audit or Off for each direction), the response options, rule sets, bypass paths and the service's WAF exclusions. How the WAF works for this service shows the current modes as a flow and says in one sentence what they do. See Turn the WAF on for a service.
  • On the Access tab, Authentication requires sign-in. See Authentication.

The WAF tab: requests in Block, responses in Audit

Response compression

Compress Responses (in the service's Compression section, off by default) makes NullWard compress responses for visitors whose browsers accept it: Brotli when the browser accepts it at least as readily as gzip, otherwise gzip. It needs NullWard 0.6.90 or later.

  • What's compressed: text-based responses (HTML, CSS, JavaScript, JSON, XML, SVG, fonts and similar) of 1 KB or more, or of unknown length.
  • What passes through unchanged: responses your backend already compressed, partial (range) responses, server-sent events and other streams, responses marked Cache-Control: no-transform, and visitors whose browsers accept neither encoding.
  • Compressed responses carry Vary: Accept-Encoding, and an ETag becomes a weak validator (W/"…").
  • The WAF inspects the response before it's compressed.
  • With the Static Asset Cache on, each encoding is cached as a separate copy, so a cached response always matches what the browser accepts.

Compressed request bodies

Decompress Request Bodies (in the same Compression section, off by default) lets the service accept request bodies sent compressed with gzip, deflate or Brotli: the WAF and the malware scan inspect them decompressed, and the backend receives them unchanged. Without it, the WAF blocks compressed request bodies. See Web application firewall. It needs NullWard 0.6.91 or later.

gRPC

Allow gRPC (in the service's gRPC section, off by default) lets the service proxy gRPC calls, including server-streaming and bidirectional streaming calls, to its destinations or through an application tunnel. It needs NullWard 0.6.93 or later. Without it, the WAF blocks gRPC: the CRS doesn't allow the application/grpc content type, and the binary request body trips its checks.

With Allow gRPC on, a gRPC request (Content-Type: application/grpc):

  • goes to the destination over HTTP/2: over TLS to an https destination (also with TLS verification skipped), or as cleartext HTTP/2 (h2c) to an http destination, which is what a gRPC server's plain port speaks;
  • is inspected by the WAF by its path, method and headers, and message by message as it streams (NullWard 0.6.94 or later). Each message is decompressed and its protobuf decoded without the service's schema; every text field it holds is checked by the WAF rules as a form argument named by its field numbers (for example f.3.1), and by the ML WAF when ML is on for the service. The content-type restriction is lifted for gRPC requests only. The malware scan skips them;
  • streams in both directions as it arrives, and the read timeout only limits the wait for the first response, so long-lived streams aren't cut.

A refused message never reaches the backend. If the backend hasn't answered yet, the client gets the gRPC status PERMISSION_DENIED with the block reference; otherwise the stream is reset. Either way a WAF (or ML) event is recorded, and audit mode only logs. A message NullWard can't decode (compressed with a codec other than gzip or deflate, or larger than 16 MiB) is refused with INVALID_ARGUMENT.

Field names are field numbers, not the names in your .proto files, so a false positive can't be excluded for one field. A whitelist rule for the call's path, which is /package.Service/Method, and the rule IDs excludes it for that method. Numeric and binary fields aren't inspected.

Other requests to the service are handled exactly as before. Through an application tunnel, gRPC needs connector 0.6.93 or later (see Application tunnels).

TLS passthrough

A TLS passthrough service forwards connections to its hostnames without decrypting them. Use it for SSL VPN gateways and other TLS services that need their own certificate or end-to-end encryption to the client. It needs NullWard 0.6.95 or later.

Choose TLS passthrough as the Upstream Type when you add the service, or as the Service Type in the service's General section, and set:

  • Connect To: Direct when NullWard can reach the target itself, or Application Tunnel for a target behind a tunnel (connector 0.6.96 or later).
  • Target (direct): the host:port connections are forwarded to, for example 10.0.0.5:443.
  • Tunnel and Upstream Group (application tunnel): the connector connects to a target of that group. The group's protocol doesn't matter, because the encrypted bytes are relayed as they are.
  • PROXY Protocol (optional): v1 (text) or v2 (binary) sends the client's address to the target in a PROXY protocol header before the TLS data. Turn it on only when the target is configured to accept PROXY protocol from NullWard; otherwise its handshakes fail.

NullWard reads the hostname from each connection's TLS hello (SNI) on its HTTPS port. A connection for a passthrough hostname is passed to the target as it is, and the client sees the target's certificate. NullWard holds no certificate for these hostnames. Every other connection is handled as before. Clients must send the hostname in the TLS hello, which browsers and VPN clients do when they connect by name. A plain HTTP request to a passthrough hostname gets a 404.

NullWard can't see inside the connection, so only the IP-based checks apply, once per connection, before it is forwarded:

CheckApplies as
MaintenanceRefuses connections while maintenance mode is on in block mode, or during a full-block maintenance window.
IP AllowlistRefuses clients outside the service's allowlist.
BlocklistRefuses blocked IP addresses.
IP IntelligenceRefuses clients the check would block. Without HTTP there is no challenge page, so a challenge also refuses.
Rate LimitingCounts connections, not requests. The limit is the higher of the per-minute rate and the burst. A VPN client usually holds one long connection, so a low limit only affects reconnects.

A refused connection is closed without an answer, and the reason is logged (TLS passthrough connection refused). The service's statistics count one request per connection, plus the bytes sent each way. The WAF, ML WAF, authentication, bot detection and the HTTP settings don't apply, so the service page hides them.

Through a tunnel, an older connector isn't used: the connection is refused and the log says to update the connector. An http health check on the group sends plain HTTP to the TLS port and marks the target down, so for a passthrough group use protocol: https with a path the target answers, or no health check (see Application tunnels).

Other sections

These sections are on the Access, Threat protection, TLS and Traffic & limits tabs.

SectionWhat it does
TLS ConfigurationCertificates, TLS versions and ciphers, HTTP/3, upstream TLS verification, HTTPS redirect and HSTS. See TLS certificates.
Bot DetectionChallenge visitors with Turnstile, reCAPTCHA or hCaptcha. See Threat protection.
IP IntelligenceCheck visitor IPs against a reputation provider. See Threat protection.
IP AllowlistAllow only listed IPs to reach this service. All other IPs get a 403. See Threat protection.
ML SecurityML WAF settings for this service (licensed). See ML WAF.
Malware ScanningScan uploads with ClamAV. See Threat protection.
WAF ExclusionsWhitelist rules for this service. See Web application firewall.
Path Access RulesDeny a path to everyone, or allow it only to certain groups of authenticated users.
Rate LimitingRequests Per Minute and Burst Allowance per client. Clients over the limit get HTTP 429.
Exception Auto-ApprovalHow exception requests for this service are handled.
Behavior: HTTP Error ScoringCount backend 4xx errors toward a client's behaviour score.
Queue / Load ManagementWaiting room (licensed). Set Max Concurrent Users, and further visitors wait on a queue page that shows their position.
Static Asset CacheCache static assets (JS, CSS, images, fonts) at NullWard to reduce backend load. The cache follows your application's Cache-Control, Expires and Vary headers. It never stores responses that set cookies, are marked no-store, private or no-cache, or answer requests with Authorization (unless the response allows sharing). A browser asking for a fresh copy always gets one from the backend. Responses without caching headers are kept for the configured default time.
Request Audit LoggingKeep a searchable log of recent requests for this service, viewable from the service's request-log page.
Request Size LimitsMaximum request body size for this service (empty = no limit).
Proxy & TimeoutsRead, write, idle and WebSocket idle timeouts for this service.
Upstream Forwarding HeadersWhich X-Forwarded-*, X-Real-IP and X-Request-ID headers are sent to the backend.
SessionsSign-in session length and the Secure cookie flag for this service.
Upstream Cookie SecurityForce the Secure, HttpOnly and SameSite attributes on cookies set by the backend.

Many of these have global defaults under Settings. A blank field on the service uses the global default.

Service statistics

The chart icon in the services list opens a service's statistics: requests, latency, bandwidth, blocks by type and cache use, over 24 hours, 3 days or 7 days. The Dashboard shows the same figures for all services together.