CrashLoopBackOff: find the failure before restarting anything
Separate application crashes, probe failures, completed commands, and memory kills with container-specific evidence.
Read the troubleshooting guideLCA field guides · free to read
Linux & Kubernetes.
Find the cause. Fix with confidence.
Start with the error you see. Get the checks, what they mean, and a safer path to recovery.
01 / Inspect02 / Explain03 / Verify
What are you troubleshooting?
Public answers. No account required.
10 full guides · 4 quick checks
OOMKilled and exit 137: prove which memory boundary failed
Distinguish a container memory kill from a generic SIGKILL, then choose a capacity-aware correction.
Read the troubleshooting guideImagePullBackOff: identify the registry failure, not just the retry
Read the exact pull error, check image identity and secret references, and separate node connectivity from application DNS.
Read the troubleshooting guidePending Pods and PVCs: separate scheduling from storage
Use scheduler and volume evidence to distinguish expected delayed binding, capacity constraints, and real provisioning failures.
Read the troubleshooting guideKubernetes DNS failures: separate naming, resolution, and connectivity
Check Service identity, client namespace, EndpointSlices, and DNS health before changing CoreDNS.
Read the troubleshooting guideNode DiskPressure: identify the filesystem and preserve the data
Separate free-space and inode pressure, inspect the actual node layout, and avoid destructive runtime cleanup.
Read the troubleshooting guideLinux “No space left on device”: disk space, inodes, and deleted files
Diagnose a full Linux filesystem with df, du, inode checks, and lsof. Choose a safe recovery without blindly deleting logs or application data.
Read the troubleshooting guideLinux Permission denied: fix ownership, directory access, ACLs, and SELinux
Work through Linux Permission denied in order: process identity, directory traversal, file modes, ACL masks, and SELinux. Avoid chmod 777.
Read the troubleshooting guidesystemd service failed: find the first error with systemctl and journalctl
Diagnose failed Linux services using unit state, exit status, journal logs, and effective configuration. Fix the cause before restarting.
Read the troubleshooting guideSSH Permission denied (publickey): troubleshoot keys without locking yourself out
Separate SSH timeouts, refused connections, and public-key authentication failures. Check the user, identity file, server policy, and permissions safely.
Read the troubleshooting guideQuick checks
Short starting points from the original Linux Help library.
Check whether a system uses UEFI or BIOS
Identify the active firmware mode before repairing GRUB.
Inspect · Boot
[ -d /sys/firmware/efi ] && echo UEFI || echo BIOSRun on the actual host, not inside a container or restricted mount namespace. Missing firmware information there cannot reliably identify the host boot mode. Consult your distribution before changing bootloader paths.
Investigate an interrupted DNF or YUM transaction
Read the transaction and dependency state before attempting recovery.
Inspect · Packages
sudo dnf history list
sudo dnf history info last
sudo dnf checkFor DNF-based systems such as Oracle Linux 8. Preserve the actual errors. Do not blindly undo transactions, remove packages, or approve a large dependency removal. Select the vendor-supported recovery only after reviewing the failed transaction.
Investigate a read-only filesystem
Determine why writes stopped before attempting a remount.
Inspect · Storage
findmnt -no TARGET,SOURCE,FSTYPE,OPTIONS /
sudo journalctl -k -b -p warning --no-pager -n 100Replace / with the affected mount. A read-only mount may be intentional, or it may reflect device or filesystem errors. Preserve data and arrange the filesystem-specific offline repair when required; do not force read-write access to damaged storage.
Trace a Linux DNS resolution problem
Compare the application resolver path with an explicit DNS query.
Inspect · Network
getent hosts example.com
dig example.comdig may require your distribution’s DNS utilities. getent follows the configured name-service lookup path; dig queries DNS. Inspect resolvectl status only if systemd-resolved is actually in use. One failed domain is not a reason to replace the machine’s resolver configuration.