PatchWardLinux Patching & Inventory Request an evaluation

Version 1.0 · self-hosted

Every host patched.
Every patch provable.

A single agent on each host reports what is installed and what is missing. Work is scheduled in waves; PatchWard executes them, confirms that every host has returned to service, and records what changed — package by package, host by host.

Debian · Ubuntu · RHEL · Rocky · Alma · SUSE · Arch

One agentper host, self-updating
Three containersconsole, API, database
No internetrequired to install or run
30 dayson your own infrastructure

The console

The console, screen by screen

Inventory, waves, execution history, calendar and the printable report.

Capabilities

Know, act, verify

Inventory of every enrolled host

Know

Inventory that reflects the current state

Hardware, operating system, kernel, packages, pending updates and systemd services for every enrolled host. Refreshed on a schedule and again after each patch run, so the data acted upon is the state of the machine rather than a manually maintained record.

Patch waves

Act

Structured patch campaigns

Hosts are grouped by environment, application or organization unit, with scope set to all available updates or security updates only. Waves can be scheduled once, weekly or monthly, suspended during a change freeze, and gated behind the four-eyes principle, which requires a second person to approve the run before any package is installed. A wave triggered twice does not patch the same host twice.

Per-server report for a wave

Verify

A reboot is recorded when the host returns

Not at the point the command is issued. PatchWard waits for the agent to report a new boot id, then records the time of return and the uptime since. Outstanding reboots remain visible until they are completed.

Governance

Authority and exceptions, both on the record

Patching touches production systems, so both questions carry a documented answer.

Users, roles and application scope

Access control

Five roles, from read-only to full control

Viewer has read and print access only. Application user sees only the servers belonging to the applications they own and may request an exception. Operator creates and executes waves. Admin approves runs and administers users. Superadmin additionally controls company branding and licensing.

Every account is scoped to the applications it belongs to: a team lead opens the console and sees the systems under their responsibility, not the entire estate.

Exception requests and their approval trail

Exceptions

Exceptions with an owner, a reason and an expiry

The application owner submits the request with a justification — a change freeze, a ticket reference, a month-end close — and an administrator approves or rejects it. Once approved, the host is skipped on its next run and the exception is consumed, so it cannot become a permanent exclusion that no one remembers granting.

The requester, the justification, the approver and the run that consumed the exception are all retained, and the skipped host is reported as an exception rather than counted as patched.

Operation

The operating cycle

The same five steps in every cycle, from five hosts to several thousand.

  1. 1

    Onboard

    The console connects to the host over SSH, installs the agent and begins receiving reports within a minute, with the log streamed live.

  2. 2

    Scan

    Each agent reports inventory and pending updates on its own cadence, or on demand when an immediate refresh is required.

  3. 3

    Schedule

    Hosts are assigned to a wave with a defined scope and window — test tier on Tuesday, production on the second Saturday — in line with the change process.

  4. 4

    Run

    The wave dispatches one job per host, each with a live console and per-host outcomes as they complete. Kernel CVEs can be applied without a reboot through kpatch on RPM hosts.

  5. 5

    Report

    Coverage, failures, pending reboots and the exact version changes, printable for the auditor, the service owner or the next engineer on call.

Printable wave report

Evidence

The record an audit requires

Every execution is recorded and printable: the host, the package, the version before and after, who triggered the run and who approved it. Coverage and success rate for the window, the hosts that failed, and the reboots still outstanding.

  • Compliance report across any date range
  • One report per wave, or per individual execution
  • Your company name in the header, PDF from the browser

Evaluate it on your own infrastructure

No prepared demonstration data. The software is installed on a machine you control, onboarded against your own hosts and exercised in a real patch window. Provide the approximate host count and the distributions in use, and the installation bundle is sent to you.

Request an evaluation

leandro.gomes@patchward.net · no sales call, no form, no account

Common questions

The four asked most often

What has to be installed?

Three containers on a single machine — console, API and PostgreSQL — from pre-built images. There is no build step and no internet access is required on that machine. Managed hosts receive a single agent, deployed from the console over SSH.

Does anything leave my network?

No. Agents communicate only with your own server. There is no telemetry, no outbound reporting and no account to create.

Which distributions are supported?

Debian and Ubuntu, RHEL, Rocky, Alma and CentOS Stream, SUSE, and Arch. Kernel live-patching through kpatch on the RPM family.

How is it removed?

Stop the three containers and drop the volume. On each host the agent is a single systemd service and one directory, and the console can remove it.