About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
A console line such as A start job is running for /dev/disk/by-uuid/... (1min 5s / 1min 30s) is not an error. It is a progress line: systemd started a job, the job has not finished, and the line shows how long it has waited and the limit. The unit named is what the boot is waiting for. Examples are a disk or partition that is not there, a mount that cannot complete, or a service waiting for the network.
Waiting is a reasonable first action. When the limit passes, systemd reports a timeout, and the journal then says which unit it was. Work from that, not from guesswork.
What the symptom means
The text comes from the service manager's status printer. In the source, the line is "%sA %s job is running for %s (%s / %s)", with the job type (start for a boot job, stop at shutdown), the unit's name or description, the time waited so far, and the limit. When the job has no timeout the limit is printed as the string no limit. A job counter, for example (2 of 3) , is added when several jobs are running.
The wording depends on how the system manager is configured. The StatusUnitFormat= setting in systemd-system.conf chooses between unit names, descriptions or both. In the source, only the description format prints A start job is running for .... The other formats print Job <unit>/<type> running (<time> / <limit>) instead. So the same wait can look different between distributions.
The limit is the unit's timeout. The systemd-system.conf manual documents DefaultTimeoutStartSec= and DefaultDeviceTimeoutSec=, and says both default "to 90 s in the system manager". A device wait is the second one: for an /etc/fstab entry, x-systemd.device-timeout= sets "how long systemd should wait for a device to show up before giving up". This is why a missing disk can stop a boot for about a minute and a half.
no limit is not a hang. The systemd.mount manual says an NFS bg mount is handled as if "x-systemd.mount-timeout=infinity,retry=10000" were prepended, which makes the wait unlimited.
When the timeout passes, the source shows what is logged: for a device, Timed out waiting for device %s., and for units that depended on it, Dependency failed for %s. A local mount that fails then holds back local-fs.target, and "Linux VM in Emergency Mode After a Disk Change: fstab Checks" covers what follows. The systemd.mount manual says a mount gets that dependency "unless one or more mount options among nofail, x-systemd.wanted-by=, and x-systemd.required-by= is set".
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
A disk, partition or UUID in /etc/fstab that does not exist |
A start job is running for /dev/disk/by-uuid/... or a .device unit; then Timed out waiting for device |
The UUID in /etc/fstab against what the VM now has; lsblk -f after boot |
| A network mount with the network not ready | Job for a mount or network-online.target |
A remote file system in /etc/fstab; systemd-networkd-wait-online or NetworkManager wait service in the waiting list |
An NFS mount with bg |
A long or unlimited wait. This is a reading of two sources: systemd.mount treats bg as if x-systemd.mount-timeout=infinity,retry=10000 were prepended, and the status printer prints no limit for a job with no timeout; neither page says the line shows no limit for a bg mount |
The bg option in /etc/fstab |
| A service that waits for the network or a slow dependency | A named service with a time and limit | systemctl list-jobs after login; journal for that unit |
| Disk or controller changed after a migration or reconfiguration | Waiting for the root or data device | Device names or UUIDs changed; see "Linux boot failure after migration: drivers, identifiers and network names" |
Prove the cause before changing anything
If you can wait, let the limit run out and note the unit name on the console. If you can reach the VM, from the console shell or by SSH after the boot, read:
systemctl list-jobs
journalctl -b
systemctl --failed
findmnt --verify
systemctl list-jobs lists jobs in progress. journalctl -b shows the current boot, where you can search for Timed out waiting for device and Dependency failed. systemctl --failed lists units that failed. findmnt --verify checks whether /etc/fstab can be parsed and used.
systemd-analyze is useful after the boot, but know its limits. systemd-analyze blame shows the time units spent in "activating" state, which "is not defined for units such as device units that transition directly from "inactive" to "active"", so the command "cannot accurately reflect latency introduced by waiting for hardware and similar events". systemd-analyze critical-chain "does not show information on jobs (and in particular not jobs that timed out)". A boot that waited for a missing disk can look innocent in both. Use the journal for timeouts, and use critical-chain for the order of units that did start.
systemd-analyze blame
systemd-analyze critical-chain
Example (illustrative): db01 was cloned and its data disk replaced, so the old UUID is gone. At boot the console shows A start job is running for /dev/disk/by-uuid/0a1b2c3d-... (45s / 1min 30s). After the 90 seconds (1 min 30 s = 90 s) the wait ends and the boot does not complete. journalctl -b shows Timed out waiting for device for that UUID, then Dependency failed for the mount. In /etc/fstab the line for /data still has the old UUID. lsblk -f shows the new disk with a different UUID.
Fix
A device that is not there
Fix the entry. Either correct the identifier to the real disk (see lsblk -f and the blkid tool), or, if the disk is optional, add nofail and a shorter x-systemd.device-timeout=. The fstab(5) manual defines nofail as "do not report errors for this device if it does not exist." In systemd terms the mount is then only "wanted, not required" by local-fs.target.
Risk: an /etc/fstab edit can stop the next boot if it is wrong. Copy the file first, run findmnt --verify after editing, and restore the copy to back out. nofail also means the system boots without that mount, and an application may then write to the empty directory underneath. See "Linux VM in Emergency Mode After a Disk Change: fstab Checks".
If the VM is at the emergency prompt and the root file system is read-only, follow the rescue guidance in "Getting into a Linux server you cannot log in to: rescue targets, rd.break and boot loader passwords" before editing.
Network mounts
Mount units for remote file systems "automatically pull in" network-online.target and order after it. The systemd.special manual says network-online.target pulls in a service that "possibly adds substantial delays to further execution". For the network wait itself, systemd-networkd-wait-online documents --timeout=SECS, "A timeout of 0 disables the timeout. Defaults to 120 seconds", and the NetworkManager service delays the target "until NetworkManager reports that the startup is completed on the D-Bus".
For a mount that is not essential at boot, the systemd.mount manual suggests x-systemd.automount, which creates an automount unit and mounts on first access, instead of NFS bg. Add _netdev if the mount type is not recognised as a network mount.
Risk: moving a remote mount to x-systemd.automount or adding _netdev in /etc/fstab changes when and how the mount happens, and a wrong line can stop the next boot or leave the path empty until first access, so an application may write to the empty directory underneath. Copy /etc/fstab first, run findmnt --verify after editing, and restore the copy to back out.
A service that waits
Read the unit with systemctl cat <unit>, then read why it is waiting in journalctl -b -u <unit>. If it needs the network and is not essential to boot, ask whether it should be ordered after network-online.target at all. The NetworkManager manual says that "Optimally, all services on the host react dynamically to network changes and systemd services do not need to be configured to start after reaching the network-online target."
Disruptive step: changing a unit's ordering or timeout can change what starts first. Make the change in a drop-in file, test with a reboot in a window, and remove the drop-in to undo.
Confirm the fix
- The next reboot shows no
A start job is running forline, or the wait ends in a few seconds. journalctl -bhas noTimed out waiting for deviceorDependency failedlines.systemctl --failedis empty, andfindmnt --verifyreports nothing.- The mount or service in question is present after boot.
What to watch afterwards
- Boot time. Compare
systemd-analyze critical-chainbefore and after, remembering what it omits. - After any disk change. Check
/etc/fstabidentifiers when a disk is replaced, cloned or migrated. - Network dependence. Keep remote mounts out of the critical path where possible.
- Optional mounts. Document which mounts carry
nofail, so a missing mount is noticed by monitoring and not by an application error.
Read next
Related articles you may want next.
- TroubleshootingLinux VM in Emergency Mode After a Disk Change: fstab ChecksWhy a systemd Linux VM stops at an emergency shell after a disk change, how to find the failing fstab entry read-only, and when nofail is the right choice.Read article
- TroubleshootingLinux boot failure after migration: drivers, identifiers and network namesDiagnose Linux migration failures by boot stage, installed kernel, storage identifiers and network identity before planning distribution-specific repairs.Read article
- Backup & RecoveryGetting into a Linux server you cannot log in to: rescue targets, rd.break and boot loader passwordsPlanning console access to a locked-out Linux server: rescue and emergency targets, rd.break, debug shells, and boot loader passwords.Read article
Risks and limits
- These commands have not been run or lab-tested. They come from the systemd source and manual pages listed below.
- The status line format is from the current systemd source. Older versions printed it differently, and distributions can change the default format.
- The default timeouts are 90 seconds in the manual pages; units, mounts and distribution builds can override them.
- The sources do not describe encrypted volumes, LVM activation waits or storage multipath, which also produce long waits at boot.
nofailhides a missing mount at boot. It is a decision about the data, not only about the boot.
Sources
- systemd src/core/manager.c (the status line) — systemd project, no date shown, accessed 18 November 2026
- systemd src/core/device.c (device timeout message) and src/core/job.c (dependency message) — systemd project, no date shown, accessed 18 November 2026
- systemd-system.conf(5) — man7.org, no date shown, accessed 18 November 2026
- systemd.mount(5) — man7.org, no date shown, accessed 18 November 2026
- systemd.special(7) — man7.org, no date shown, accessed 18 November 2026
- systemd-analyze(1) — man7.org, no date shown, accessed 18 November 2026
- systemctl(1) and journalctl(1) — man7.org, no date shown, accessed 18 November 2026
- systemd-networkd-wait-online.service(8) — man7.org, no date shown, accessed 18 November 2026
- NetworkManager-wait-online.service(8) — Debian manpages, NetworkManager 1.42.4, no date shown, accessed 18 November 2026
- fstab(5) — man7.org, no date shown, accessed 18 November 2026
- findmnt(8) — Debian manpages, util-linux 2.38.1-5, no date shown, accessed 18 November 2026
This article was originally published on .