Plex Server won’t start

[fizzyhd]

this is the log im getting from my unraid container

2026-08-27 23:35:47.144458 [info] System information: Linux MovieServer 6.12.54-Unraid #1 SMP PREEMPT_DYNAMIC Tue Oct 21 15:58:46 PDT 2025 x86_64 GNU/Linux
2026-08-27 23:35:47.165396 [info] Image architecture: 'amd64'
2026-08-27 23:35:47.185532 [info] Application name: 'plexpass'
2026-08-27 23:35:47.205784 [info] Base image release tag: '2026072801'
2026-08-27 23:35:47.225442 [info] Application image release tag: '1.43.4.10903-1-01'
2026-08-27 23:35:47.250312 [warn] PUID not defined (via -e PUID), defaulting to '99'
2026-08-27 23:35:47.272762 [info] Executing usermod for PUID '99'...
2026-08-27 23:35:47.364217 [info] usermod completed successfully
2026-08-27 23:35:47.387608 [warn] PGID not defined (via -e PGID), defaulting to '100'
2026-08-27 23:35:47.411095 [info] Executing groupmod for PGID '100'...
2026-08-27 23:35:47.522619 [info] groupmod completed successfully
2026-08-27 23:35:47.542608 [warn] UMASK not defined (via -e UMASK), defaulting to '000'
2026-08-27 23:35:47.565628 [info] Permissions file '/config/perms.txt' exists, skipping setting ownership and permissions on '/config' and '/data'
2026-08-27 23:35:47.595204 [info] Deleting files in /tmp (non recursive)...
2026-08-27 23:35:47.621675 [warn] TRANS_DIR not defined,(via -e TRANS_DIR), defaulting to '/config/tmp'
2026-08-27 23:35:47.717065 [info] Starting Supervisor...
2026-08-27 23:35:47,884 INFO Included extra file "/etc/supervisor/conf.d/plexmediaserver.conf" during parsing
2026-08-27 23:35:47,884 INFO Set uid to user 0 succeeded
2026-08-27 23:35:47,889 INFO RPC interface 'supervisor' initialized
2026-08-27 23:35:47,889 INFO supervisord started with pid 19
2026-08-27 23:35:48,892 INFO spawned: 'claim-script' with pid 89
2026-08-27 23:35:48,894 INFO spawned: 'start-script' with pid 90
2026-08-27 23:35:48,894 INFO reaped unknown pid 20 (exit status 0)
2026-08-27 23:35:48,908 DEBG 'claim-script' stdout output:
[info] Env var 'PLEX_CLAIM' value not set and/or 'CLAIM_SERVER' not set to 'yes', skipping claim process.

2026-08-27 23:35:48,908 INFO success: claim-script entered RUNNING state, process has stayed up for > than 0 seconds (startsecs)
2026-08-27 23:35:48,908 DEBG fd 9 closed, stopped monitoring <POutputDispatcher at 22537931527840 for <Subprocess at 22537932815536 with name claim-script in state RUNNING> (stdout)>
2026-08-27 23:35:48,908 DEBG fd 11 closed, stopped monitoring <POutputDispatcher at 22537931446608 for <Subprocess at 22537932815536 with name claim-script in state RUNNING> (stderr)>
2026-08-27 23:35:48,909 INFO exited: claim-script (exit status 0; expected)
2026-08-27 23:35:48,909 DEBG received SIGCHLD indicating a child quit
2026-08-27 23:35:48,914 DEBG 'start-script' stdout output:
[info] Env var 'PLEX_CLAIM' value not set and/or 'CLAIM_SERVER' not set to 'yes', skipping edit of Preferences.xml for claim process.
[info] Starting Plex Media Server...

2026-08-27 23:35:49,916 INFO success: start-script entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2026-08-27 23:35:50,592 DEBG 'start-script' stderr output:
Plex Media Server is already running. Will not start...

2026-08-27 23:35:50,956 DEBG 'start-script' stdout output:
[info] Plex Media Server stopped

2026-08-27 23:35:50,956 DEBG fd 12 closed, stopped monitoring <POutputDispatcher at 22537931447568 for <Subprocess at 22537931441488 with name start-script in state RUNNING> (stdout)>
2026-08-27 23:35:50,956 DEBG fd 16 closed, stopped monitoring <POutputDispatcher at 22537931399872 for <Subprocess at 22537931441488 with name start-script in state RUNNING> (stderr)>
2026-08-27 23:35:50,956 INFO exited: start-script (exit status 0; expected)
2026-08-27 23:35:50,957 DEBG received SIGCHLD indicating a child quit```

[alacran669]

##My server wont start but if I go back to last release its working normally.
[info] Env var 'PLEX_CLAIM' value not set and/or 'CLAIM_SERVER' not set to 'yes', skipping edit of Preferences.xml for claim process.
[info] Starting Plex Media Server...

2026-09-04 11:20:05,507 INFO success: start-script entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2026-09-04 11:20:21,782 DEBG 'start-script' stdout output:
Error: Unable to set up server: sqlite3_statement_backend::loadOne: database or disk is full (N4soci10soci_errorE)

2026-09-04 11:20:22,036 DEBG 'start-script' stdout output:
[info] Plex Media Server stopped

2026-09-04 11:20:22,036 DEBG fd 12 closed, stopped monitoring <POutputDispatcher at 23049009716496 for <Subprocess at 23049009710416 with name start-script in state RUNNING> (stdout)>
2026-09-04 11:20:22,036 DEBG fd 16 closed, stopped monitoring <POutputDispatcher at 23049009668800 for <Subprocess at 23049009710416 with name start-script in state RUNNING> (stderr)>
2026-09-04 11:20:22,036 INFO exited: start-script (exit status 0; expected)
2026-09-04 11:20:22,037 DEBG received SIGCHLD indicating a child quit

[jonoff]

Plex Media Server is already running. Will not start...

This is the clue, find and delete the existing plexmediaserver.pid file.
rm /<config>/Plex\ Media\ Server/plexmediaserver.pid

[jhp612]

Hitting the same thing on Unraid with 1.43.4.10903-1-01 (base 2026072801), and I dug into why. On my box it happens after every host reboot, and a plain container restart always fixes it. Here's what I found.

Root cause: stale plexmediaserver.pid from an unclean stop, plus deterministic PIDs

1. PMS never gets a clean shutdown. On docker stop, supervisord sends SIGTERM to start-script, but PMS doesn't appear to receive it. Nothing is logged in Plex Media Server.log, and supervisord just repeats waiting for start-script to die until its default stopwaitsecs (10s) runs out. Then it SIGKILLs:

   INFO waiting for start-script to die
   INFO waiting for start-script to die
   INFO waiting for start-script to die
   INFO waiting for start-script to die
   WARN killing 'start-script' (114) with SIGKILL
   

   Raising the Docker stop timeout (I tried docker stop --time=90) doesn't help, because supervisord's own 10s limit fires first. By contrast, kill -TERM <pms pid> sent directly makes PMS exit cleanly in about 0.5s and remove its pid file. So the signal isn't reaching PMS through start.sh. I haven't pinned down the exact mechanism, but it lines up with start.sh no longer execing PMS (b67c2fb) and waitproc.sh trapping SIGTERM in the parent bash (fcaa415).

2. So /config/Plex Media Server/plexmediaserver.pid is left behind on every stop.

3. Why it breaks only after a reboot: PIDs inside a fresh container are almost deterministic. In my logs, a restart spawns start-script at pid 115 and PMS lands on 119, which gets written to the pid file. After a host boot, start-script is pid 113/114, so PMS lands one or two lower (118). Its first thread then gets TID 119, which is the stale number in the file (TIDs share the PID space and /proc/<tid> exists). PMS sees a live "Plex Media Server" at 119 that isn't itself and refuses with already running. As far as I can tell, PMS's check is "does this pid exist and look like Plex". A pid file containing 1 did not block startup, which is why this looks random and why a restart (different PID layout) makes it go away. The container still shows healthy because the healthcheck watches supervisord, not PMS, and nothing retries.

Workaround (same approach linuxserver has used since 2019): delete the pid file before PMS starts. linuxserver/docker-plex does this in init-plex-chown/run ("remove plex pid after unclean stop", fix for their #134). On Unraid, I put this in the container's Post Arguments field, which overrides the image CMD (/bin/bash init.sh):

/bin/bash -c 'rm -fv "/config/Plex Media Server/plexmediaserver.pid"; exec /bin/bash init.sh'

Outside Unraid, append the same thing after the image name in docker run. With this in place, the container log prints removed '/config/Plex Media Server/plexmediaserver.pid' on start and PMS comes up normally.

Suggested fix in the image (either or both):

• rm -f "/config/Plex Media Server/plexmediaserver.pid" in start.sh before launching PMS. It's always stale at that point, since only one PMS can run per container.
• Make SIGTERM actually reach PMS (e.g. exec it again, or have the trap forward TERM to the PMS pid) and/or raise stopwaitsecs in plexmediaserver.conf. That way PMS shuts down cleanly and isn't SIGKILLed every stop, which also means SQLite is closed properly instead of recovering the WAL on every start.