Why Did My Server Go Down? How to Find the Cause of an Outage
A server rebooted or dropped off the network. The logs and commands that show whether it ran out of memory, crashed, lost a disk or had its network changed.
By Frabs · 15 November 2026 · 4 min read
The short answer: start with the logs from the boot before the outage (journalctl -b -1), then look for the out-of-memory killer, kernel panics, hung tasks and disk errors, and check whether the server shut down cleanly or simply stopped. Most outages fall into a handful of causes, and each leaves its own trace.
When a customer asks why their server went down, "it rebooted" is not an answer. The useful answer says what made it reboot, and whether it will happen again.
First: Did It Reboot or Lose Its Network?
These are different problems. Check how long the server has been up:
uptime -sIf the boot time is just after the outage, the server restarted. If it has been up for weeks, the server stayed up and something cut it off from the network.
To see recent boots and shutdowns:
last -x reboot shutdown | headA reboot with no shutdown just before it means the server stopped without shutting down: a crash, a power loss or a hard reset.
Read the Logs From Before the Outage
If the journal is kept on disk, the previous boot's log is still there. Warnings and errors from the end of it are where the cause usually is:
journalctl -b -1 -p warning -n 200If that says there is no previous boot, the journal is only kept in memory. Make it persistent for next time:
mkdir -p /var/log/journal && systemctl restart systemd-journaldThe Usual Causes and What They Look Like
Out of memory. The kernel killed processes to free memory, and sometimes the server became unresponsive before it recovered:
journalctl -b -1 -k | grep -iE 'out of memory|oom-kill'A kernel panic or crash. The kernel stopped. Some servers keep a record of the crash in pstore, which survives the reboot:
ls /sys/fs/pstore/ /var/lib/systemd/pstore/ 2>/dev/nullHung tasks. Processes stuck waiting, usually on storage, often just before a server stops responding:
journalctl -b -1 -k | grep -i 'blocked for more than'Disk errors. I/O errors, a failing drive or a degraded RAID array:
journalctl -b -1 -k | grep -iE 'i/o error|medium error|ata[0-9]+.*error'
smartctl -H /dev/sdaHeat or power. Thermal throttling and machine check errors show in the kernel log, and IPMI's event log often records power events the operating system never saw:
ipmitool sel list | tailA planned or clean reboot. A shutdown entry just before the boot, often from an update or someone running reboot. Check your package manager's log and shell history before looking further.
A network change. The server stayed up, but an address, route, bridge, bond or VLAN changed and cut it off. Look for interface and address changes in the log around that time:
journalctl --since '-2h' | grep -iE 'link (is )?(down|up)|carrier|netplan|ifup|ifdown'Keep the Evidence for Next Time
Most of the time the cause is found only because something was recorded before the outage. Keep the journal persistent, keep IPMI's event log, and keep resource history (CPU, memory and disk over time), so that a slow climb in memory is visible before the crash rather than guessed at afterwards.
How Frabs Tells You Why
On Frabs Host and Frabs Hypervisor, each server reports every minute. When one goes down, the outage on its uptime record says why: it ran out of memory, the kernel crashed, a disk failed, it overheated, the power was cut, a network change cut it off, or it was simply rebooted. Show Logs opens the server's own kernel and system log from just before it went down, so you do not have to go looking.
Frabs only records downtime it can prove. If the sensor stopped but the server stayed up, that is not shown as an outage. The reasons and logs stay in your control panel, and the public status page you give customers shows only the up and down timeline.
Frequently Asked Questions
How do I see why my Linux server rebooted?
Read the previous boot's log with journalctl -b -1 and look for the out-of-memory killer, kernel panics, hung tasks and disk errors. last -x shows whether the server shut down cleanly before it booted again.
Why is there no log from before the reboot?
The journal may only be kept in memory, so it is lost at reboot. Create /var/log/journal and restart systemd-journald to keep it on disk.
How can I tell a crash from a power loss?
A kernel crash often leaves a record in pstore and messages in the previous boot's log. A power loss usually leaves nothing in the operating system's logs, but may show in IPMI's event log.
Does Frabs show why a server went down?
Yes, on Frabs Host and Frabs Hypervisor. Every outage says why it happened, and Show Logs opens the server's own kernel and system log from just before. This stays in your control panel and is never shown on public status pages.