OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
network medium very-common DISPUTE

Stop NetworkManager-wait-online adding 30 to 120 seconds to every boot

omarchyarchendeavouroscachyosmanjarolaptopdesktop
CORRECTED An auditor found the fix wrong or incomplete and rewrote it.
AUDIT NOTE

The problem, the remedy and every Omarchy-specific claim are correct, but one sentence of mechanism inside the fix is false, and it is exactly the confident-specific failure mode. Verified true: Omarchy 4's install/config/enable-services.sh on quattro really does systemctl mask NetworkManager-wait-online.service, migrations/1784568652.sh really does mask it on upgrade, and install/hardware/network.sh really does disable+mask systemd-networkd-wait-online.service while retiring archinstall's 'copy ISO network' units. The record even gets the archinstall provenance right. The cups-browsed cause is confirmed by the migration's own comment: "graphical.target was gated on network-online.target (cups-browsed orders itself after it)". The unit file on this machine confirms ExecStart=/usr/bin/nm-online -s -q, Environment=NM_ONLINE_TIMEOUT=60 and WantedBy=network-online.target, so dropping -s to wait for real connectivity and raising NM_ONLINE_TIMEOUT are both right. What is wrong: "Mask, not disable: systemctl disable will not stop it, because it is pulled in as a dependency of network-online.target rather than started on its own." The enablement symlink is /etc/systemd/system/network-online.target.wants/NetworkManager-wait-online.service (there is no vendor .wants directory under /usr/lib), so systemctl disable does remove it and does stop it running. The real reason to prefer mask is in NetworkManager.service's own [Install]: Also=NetworkManager-wait-online.service, so any later systemctl enable NetworkManager or preset run silently re-enables it. Corrected fix replaces that one rationale with the verified one. All commands are unchanged.

The cause above was not rewritten and may still contain the error described. The fix below is the corrected version.
RISK

Masking the unit means network-online.target is reached immediately, so any unit ordered After=network-online.target (NFS/CIFS mounts in fstab with _netdev, wg-quick@, backup timers, self-hosted services) can now start before there is an address and fail on the first try. Audit systemctl list-dependencies network-online.target before masking, and give anything genuinely network-dependent its own Restart=on-failure / RestartSec= rather than relying on the global wait.

Symptom

Boot sits on a blank screen or the splash for half a minute or more before the login screen appears, sometimes ending with A start job is running for Network Manager Wait Online (1min 30s / no limit) or a red Failed to start Network Manager Wait Online. systemd-analyze blame puts NetworkManager-wait-online.service at the top with 30s, 60s or 120s. It is worst on laptops that boot away from their usual Wi-Fi, and on machines with no cable plugged in.

Cause

NetworkManager-wait-online.service is WantedBy=network-online.target, so it only runs when something pulls that target in, and on a desktop install something almost always does. On Omarchy it is cups-browsed.service, which orders itself after network-online.target, which in turn gates graphical.target. The result is that the whole desktop waits for DHCP or Wi-Fi association before it will draw. Nothing in a Hyprland session actually needs the network to be up before it starts.

Fix

First confirm it is the culprit:


systemd-analyze blame | head -10
systemd-analyze critical-chain graphical.target
systemctl list-dependencies network-online.target --reverse

Then mask the unit:


sudo systemctl mask NetworkManager-wait-online.service

**Mask rather than disable.** systemctl disable does work, since the enable symlink lives in /etc/systemd/system/network-online.target.wants/ and disabling removes it, but it does not stick. NetworkManager.service carries Also=NetworkManager-wait-online.service in its own [Install] section, so the next systemctl enable NetworkManager.service, or any preset run, quietly re-enables the wait unit. A mask survives all of that.

If systemd-networkd is also installed (common on machines built by archinstall's "copy ISO network config" mode), mask its equivalent too:


sudo systemctl mask systemd-networkd-wait-online.service

Omarchy 4 does both of these for you: install/config/enable-services.sh masks the NetworkManager one on a fresh install, install/hardware/network.sh disables and masks the networkd one, and migrations/1784568652.sh masks the NetworkManager one on upgrade, so on Omarchy check first rather than assuming:


systemctl is-enabled NetworkManager-wait-online.service   # expect: masked

If you genuinely have a service that must not start before the network is really up (a network mount, a VPN, a backup job), do not mask the unit. Instead make it wait for actual connectivity rather than for NetworkManager to finish starting, by dropping the -s flag (the stock unit is ExecStart=/usr/bin/nm-online -s -q, and -s means "wait until NetworkManager logs startup complete", not "wait until there is an address"):


sudo systemctl edit NetworkManager-wait-online.service

[Service]
ExecStart=
ExecStart=/usr/bin/nm-online -q

and raise the timeout in the same drop-in if the stock 60s is too tight:


[Service]
Environment=NM_ONLINE_TIMEOUT=120

Verify

Reboot, then systemd-analyze blame | head -5 should no longer list either wait-online unit, and systemd-analyze should report a total boot time tens of seconds shorter. systemctl is-enabled NetworkManager-wait-online.service prints masked.

Sources

networkmanager-wait-online-delays-boot