Skip to content
EncryptTechBD
Home Industries Our Work Blog Pricing About Contact Discuss Your Project
বাংলা (BN)
Website & Infrastructure 5 min read

Server patch management checklist

A working checklist for finding every server, ranking patches by risk, testing them, deploying in waves and proving the job was done.

E
EncryptTechBD · 0 comments
iMac displaying an analytics dashboard on a bright desk
Photo: Serpstat / StockSnap

Patching fails for ordinary reasons. The server inventory is out of date, the maintenance window clashes with a business process, the test environment does not resemble production, and nobody checks whether the reboot actually applied the update. The result is a fleet that looks patched in a spreadsheet and is exposed in reality.

This checklist is written for the person who owns the servers, and for the manager who needs evidence that the work happened. It covers operating systems, packages, container images, firmware and the applications that sit on top.

Start with an inventory you trust

  • Every physical server, virtual machine, cloud instance and managed appliance.
  • Operating system, version, support status and owner.
  • Installed packages, container base images, database engines and web servers.
  • Firmware and hypervisor versions, including network and storage devices.
  • Applications and plugins with their own update cycle, such as a CMS or ERP module.
  • Systems that cannot be patched and the compensating control around them.

If the inventory is maintained by hand, it is already wrong. Generate it from the hypervisor, cloud account and configuration management tool, then reconcile the differences monthly.

The 15-point checklist

Inventory and ownership

  1. Assign an owner to every server. A named person, not a team mailbox. Evidence: inventory export with owner and last review date.
  2. Record the business service each server supports. This decides the maintenance window and the escalation path. Evidence: service-to-server map.
  3. Track end-of-support dates. An unsupported operating system cannot be patched, only isolated or replaced. Evidence: support status column and a replacement plan.

Risk and timing

  1. Define severity tiers and remediation targets. A defensible starting policy is emergency for actively exploited issues, short windows for critical and high severity, and a longer cycle for low risk. Adjust to your risk appetite and write it down. Evidence: approved policy.
  2. Use a threat feed. Check the CISA Known Exploited Vulnerabilities catalog and vendor advisories, not only the scanner score. Evidence: weekly review note.
  3. Set a maintenance calendar. Publish windows per service tier and agree them with the business. Evidence: calendar with named approvers.

Testing and change control

  1. Test in a representative environment. A patch that breaks the ERP integration will not show up on a clean test box. Evidence: test result and known issues.
  2. Take a snapshot or backup before the change. Confirm the restore path, not just the backup job. Evidence: snapshot ID and restore check.
  3. Record the change. Include the package versions, the reason, the rollback step and the person approving it. Evidence: change ticket linked to the server.

Deployment and verification

  1. Deploy in waves. A small canary group first, then the rest, with a pause to watch errors. Evidence: wave plan and pause decision.
  2. Install from trusted sources. Use vendor repositories and signed packages; pin versions where reproducibility matters. Evidence: repository configuration and package signature check.
  3. Verify after reboot. Check the running kernel or package version, service health, application login and error logs. A successful install command is not proof. Evidence: verification output per server.

Emergency and exceptions

  1. Keep an emergency patch path. Pre-approved authority, an out-of-hours contact and a rollback plan for actively exploited issues. Evidence: one-page emergency runbook.
  2. Time-box every exception. Record the risk, the compensating control, the owner and the expiry date. Evidence: exception register.
  3. Report coverage monthly. Show patched, pending, failed and excepted systems, with the oldest overdue item. Evidence: one-page report to management.

Containers and appliances need their own line

Rebuilding a container image is not the same as patching a running container. If the base image is old, every deployment inherits the problem. Pin the base image, rebuild on a schedule, scan the result, and replace running containers rather than editing them. Network appliances, hypervisors and firmware often sit outside the operating system patch tool, so give them a separate owner and cadence. A vulnerability scanner that cannot see the appliance is not evidence that the appliance is safe.

Designing a patch window that people respect

A maintenance window is a business agreement, not a technical setting. Publish it at least a week ahead, name the services that will be unavailable, and give the business a way to move the window for a genuine event. Keep the window short enough that teams do not postpone it, and long enough to include verification and rollback. If a patch requires a reboot, check the startup order and the health checks before the window closes. A window that regularly overruns trains people to ignore the next one.

  • Agree the window with the service owner, not only with IT.
  • Publish the customer-facing impact in plain language.
  • Keep a named decision-maker available during the window.
  • Record the actual start and finish time, not the planned time.

Common mistakes

  • Patching only the servers that appear in the monitoring tool.
  • Treating a vulnerability scan as a patch record.
  • Running the same change in production without a canary group.
  • Forgetting the reboot, or rebooting without checking service dependencies.
  • Letting exceptions renew automatically.
  • Leaving the patch window at a time when the business cannot tolerate downtime.

Metrics

  • Percentage of systems within the policy window by severity.
  • Oldest overdue patch, in days.
  • Failed patch rate and the reason.
  • Mean time to remediate an actively exploited issue.
  • Number of open exceptions past their expiry date.
  • Percentage of container images rebuilt in the last cycle.

Where this fits in our work

Patch management is a maintenance discipline, not a one-off project. Our Managed Technology & Cybersecurity division runs the inventory, change and verification cycle for business servers; the pricing page explains how coverage is scoped. For the hardening steps that sit alongside patching, read the Linux server hardening checklist.

Sources

Services that deliver this

Want this implemented properly?

Pricing →

Every engagement starts with a written scope. If a requirement is outside it, we quote it separately before starting.

Responses (0)

No comments yet. Add the first one below.

Leave a response

Your email address will not be published. Required fields are marked *.

Keep reading

More field notes