Linux Server Administration

High load average does not always mean processors are busy. Tasks waiting on disk, network storage, locks, or other uninterruptible work can raise load while CPU utilization remains low, making a simple CPU upgrade ineffective. What to check Compare load average with CPU states, process wait channels, disk latency, filesystem health, network mount...

A renewed certificate can be valid while an application still fails because it has not reloaded, the chain is incomplete, a private key no longer matches, or the service uses a different certificate path than expected. What to investigate Check certificate dates and names, chain validation, key match, file ownership and permissions, service...

A previously accepted SSH key can stop working after a home-directory permission change, account update, key rotation, server policy change, or host replacement. A client-side error may not distinguish authentication failure from a server configuration problem. What to check Check the exact username and destination, key fingerprint, account status,...

A deployment can create files under a different user, group, or umask than the running service expects. The application may then fail to read configuration, write uploads, or rotate logs even though the release completed successfully. What to check Compare ownership and mode bits before and after deployment, inspect the deployment user and umask, and...

Zombie processes have exited but remain in the process table until their parent collects their exit status. A small number may be harmless, while a growing count can indicate a parent process or child-management problem. What to check Identify the parent process and observe whether zombie counts grow over time. Review the parent service logs and...

A Linux filesystem can switch to read-only mode after storage errors, filesystem protection, mount configuration, or an underlying virtual disk issue. Remounting it read-write without understanding the cause can risk data corruption. What to check Check kernel logs, mount options, device health, filesystem status, and recent storage or host events....

A network share can stop mounting after a reboot, credential change, DNS update, protocol restriction, or remote permission change. Applications may then write into an empty local mount directory and create confusing data gaps. What to check Check mount status, system logs, remote host resolution, share path, protocol version, credentials, network...

A server can complete a network connection while the application behind it is stalled, blocked on a dependency, or waiting for a worker. A successful TCP handshake therefore does not prove that a user request can finish. What to check Check response time at each layer, worker and connection queues, process health, dependency latency, resource pressure,...