Lychee container doesn't start, requests "a valid cache path"

[jlgarnier]

Hi all!

I deleted the previous instance of the app, so starting a fresh install. I've created a stack with Portainer, using the preferred compose file (https://github.com/LycheeOrg/Lychee?tab=readme-ov-file#docker-recommended). Just customized the port, APP_URL, APP_KEY and passwords values.

Launching the stack leads to the following error:

✅ Database port is open!
Validating and setting PUID/PGID
  User UID: 1000
  User GID: 1000
🔍 Validating permissions...
⏰ Set Permissions for Lychee folders...
✅ Permissions set securely
⏰ Set Permissions for Upload/dist folders (this may take a while)...
✅ Permissions set securely
Checking RUN_AS_ROOT setting
✅ Will drop privileges to www-data user
🧹 Clearing bootstrap cache...
🌐 Starting Lychee in web mode...
🔄 Running database migrations...
In Compiler.php line 75:
  Please provide a valid cache path.

I just can't see where to set up this path value in the compose file... 🤔

Any help would be greatly appreciated!

[ildyria]

I am not sure this docker compose is fully functional. I will need to double check.

In the meantime, I strongly recommend you use one of those two:

• https://github.com/LycheeOrg/Lychee/blob/master/docker-compose.minimal.yaml < running docker compose up on this one should start everything directly (beware the secrets are hard coded).
• https://github.com/LycheeOrg/Lychee/blob/master/docker-compose.yaml < you may need to do a few configuration, but I believe it is nicely documented.

[ildyria]

@jlgarnier I update the docker compose: #4091
The file mount were a bit too broad. Tested locally to work. Thank you for the heads up.

[jlgarnier]

Thanks Benoit, I'll update the current stack and will also check the normal docker-compose file (I hate hard coded secrets 😉)...

[jlgarnier]

Feedback on the updated compose-file: the stack now reaches the "available" phase (port correctly exposed) but

• returns a warning WARN Caddyfile input is not formatted; run 'caddy fmt --overwrite' to fix inconsistencies.; I tried to fix it but the caddy command is not found in the container...
• returns an error ERROR could not clean default/global storage., which I don't know how to fix...
• displays a weird UI as a probable consequence of the error message (see below)... Entering any value in those fields doesn't trigger any action.

I'll now check with the docker-compose.yaml file.

[ildyria]

The warning from Caddy can be safely ignored.
The second one is more worrysome.

[ildyria]

(I suffixed the containers with _test but for the rest it is literally the same docker compose as in the readme).

Here what you should see after:

[jlgarnier]

OK, it's worth another try...

[jlgarnier]

OK, finally got the instance to launch! I created my admin account and a secondary account, I can login with both. So I logged as secondary user (without admin rights), created a new album and tried to upload a whole bunch of photos and videos already on my laptop.

I got a lot of errors and the album displays incorrectly:

Thumbnails don't display, only 2 items out of 29 have been properly uploaded and the container log still informs that

 ERROR  unable to autosave config.  
 WARN  unable to get instance ID; storage clean stamps will be incomplete.  
 ERROR  could not clean default/global storage. 

Is there any access rights issue with the 3 volumes declared for the app container?

Thanks again for your help!

[ildyria]

That is very surprising. I will double check locally.

[ildyria]

Just created an album etc. No errors are being shown in the logs.

Can you check the content of the Diagnostics page?
Also can you check what the network tab shows? It seems that there are some issues with loading images, 404 maybe?

[jlgarnier]

Thanks for the quick answer!

I instinctively tried to clear the browser cache, open a new tab and point at https://pictures.mydomain.com and landed on an error page saying

If you can read me, it means that you misconfigured Lychee.
Please check that:

- your APP_URL is properly set to the correct url.
For example: APP_URL=https://lychee.example.com
- if you are working behind a reverse proxy, that TRUSTED_PROXIES is set to the forwarding ip.
For example: TRUSTED_PROXIES=*
Note that the wildcard value (*) is a very insecure option and not recommended.
- if you are working behind an apache reverse proxy, that the forwarding headers are properly set.
For example: RequestHeader set X-Forwarded-Proto https is set.
- if the CSP is blocking your assets because serving http, you need to force https.
This is done by setting APP_FORCE_HTTPS=true

This is a new issue: I'll certainly have to tweak some parameters... APP_URL is certainly OK, I didn't set any TRUSTED_PROXIES value in the compose file, so this one is missing, the service is sitting behind a Nginx proxy and I don't think I need the APP_FORCE_HTTPS=true as the secure link is with NPM, not Lychee...

However, if I point to the internal IP address, I land on the Lychee homepage. The diagnostic page seems to enter a deadloop ("Auto-diagnostic : loading" displays forever) and the network tab lists Failed to load resource: the server responded with a status of 507 (Insufficient Storage).

I also get a lot of Loading the image '<URL>' violates the following Content Security Policy directive: "img-src 'self' <URL> <URL> <URL> <URL> <URL> <URL> <URL> <URL> <URL> data: blob:". The action has been blocked.: is it related to the TRUSTED_PROXIES settings?

[ildyria]

The action has been blocked.: is it related to the TRUSTED_PROXIES settings?

Yes. As you are working behind reverse proxy (nginx), you will need to set this value.
Also do not forget to set the forwarded proto headers.

APP_FORCE_HTTPS=true is to ensure that url are generated with https links instead of http if the detection is not done properly inside Lychee.

[jlgarnier]

OK, I need some clarification:

• I must set TRUSTED_PROXIES in the compose file (as I'm using Portainer, I don't work with .env files) to point to the internal NPM IP address. What if the NPM service is running on the same machine (so same IP)?
• What about the diagnostic deadloop? Can the point above explain this too?

Thanks in advance for your explanation!

[ildyria]

I must set TRUSTED_PROXIES in the compose file (as I'm using Portainer, I don't work with .env files) to point to the internal NPM IP address. What if the NPM service is running on the same machine (so same IP)?

You can set it to * it is not ideal, but that should take care of some of the issues.

The diagnostic page seems to enter a deadloop ("Auto-diagnostic : loading" displays forever) and the network tab lists Failed to load resource: the server responded with a status of 507 (Insufficient Storage).

It seems like there is some access write issues on your setup. 507 is usually a sign that the files (usually logs) can't be created/edited. In theory, you can use the following environment variables:

PUID: "${PUID:-1000}"
PGID: "${PGID:-1000}"

Not having a .env is not an issue as you use portainer, all environment variables are read by the docker image and used downstream.

[jlgarnier]

I don't think the PUID/PGID values have a positive impact... I now get

In StreamHandler.php line 164:

  The stream or file "/app/storage/logs/daily-2026-02-16.log" could not be opened in append mode: Failed to open stream: Permission denied
  The exception occurred while attempting to log: Illuminate\Foundation\Bootstrap\HandleExceptions::closure:Illuminate\Foundation\Bootstrap\HandleExceptions::forwardsTo():257}:52 file_put_contents(): Write of 11461 bytes failed with errno=28 No space left on device

There's 256GB left...

[ildyria]

I don't think the PUID/PGID values have a positive impact... I now get

the PUID/PGID need to match the values of the folders mounted. I my case, they are left as default and inside the folders, the id are set to 33:33 (or www-data:www-data).

[ildyria]

you can also do a chmod -R 777 * on the folders that are mounted, not ideal either as it violates the principle of least amount of priviledges

[jlgarnier]

Sorry for the shortcut: the PUID and PGID values have actually been both set to www-data (33) and the host folder (/data/compose/46/lychee/logs/) gives full rights to www-data:www-data

root@omv:/data/compose/46/lychee# ls -l
total 12
drwxr-x--- 2 www-data www-data 4096 16 févr. 15:02 logs
drwxr-x--- 6 Jean-Luc     1000 4096 11 janv. 16:01 storage
drwxr-x--- 4 www-data www-data 4096 16 févr. 11:15 tmp

This is why I'm a bit lost...

[ildyria]

So weird >_<

[jlgarnier]

Hi @ildyria !

Finally got a new instance running: new server, new platform (CasaOS), new compose file and... same behavior! Here's what I found:

• Monitoring the stack deployment with Portainer reveals that the container lychee-app fails starting as it doesn't find the lychee-db container which it depends on. I had to wait until the DB container is running and healthy to manually restart the lychee-app container. This should be fixed in the compose-file.
• I created a new admin and a new user through the GUI, set up some parameters: everything's fine with the settings.
• I created a new album: OK.
• I uploaded 3 photos + 4 videos: OK.
• I clicked "Finish" and landed on the album page.
• The album homepage doesn't display any thumbnail, I have exactly what's pictured in Lychee container doesn't start, requests "a valid cache path" #4090 (comment) ...
• If I click on any photo/video, Lychee doesn't display anything but the "missing thumbnail" icon.

Thanks for any clue!

[ildyria]

WTF >_<

When you open your network tab, what do you see ?
What is the URL used to load the pictures?

[jlgarnier]

This instance is not behind any proxy yet (the previous was), so everything is fetched from the LAN...

Here's what I get when I access the homepage:

Failed to load resource: the server responded with a status of 401 (Unauthorized)
app-D9h553fY.js:6 error CustomEvent
api/v2/Timeline?date=:1  Failed to load resource: the server responded with a status of 401 (Unauthorized)
app-D9h553fY.js:6 error CustomEvent
@vue-CEoZ36M9.js:1 AxiosError: Request failed with status code 401
    at st (axios-C0Zqfgkc.js:3:1078)
    at XMLHttpRequest.w (axios-C0Zqfgkc.js:3:5900)
    at j.request (axios-C0Zqfgkc.js:5:2078)
    at async _ (Timeline-B49bzql9.js:1:12463)
    at async Timeline-B49bzql9.js:1:12544
ih @ @vue-CEoZ36M9.js:1
login:1 Autofocus processing was blocked because a document already has a focused element.
9Loading the image '<URL>' violates the following Content Security Policy directive: "img-src 'self' <URL> <URL> <URL> <URL> <URL> <URL> <URL> <URL> <URL> data: blob:". The action has been blocked.

Here's what I get when I open the album:

error CustomEvent
@vue-CEoZ36M9.js:1 AxiosError: Request failed with status code 401
    at st (axios-C0Zqfgkc.js:3:1078)
    at XMLHttpRequest.w (axios-C0Zqfgkc.js:3:5900)
    at j.request (axios-C0Zqfgkc.js:5:2078)
    at async _ (Timeline-B49bzql9.js:1:12463)
    at async Timeline-B49bzql9.js:1:12544
ih @ @vue-CEoZ36M9.js:1
:8650/login:1 Autofocus processing was blocked because a document already has a focused element.
eafd062437bbb7c7319e2e30ecd7.jpeg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
c2f0cba689647125331c7f0d6e8f.jpeg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
345e977423e9b794f9b6caeba92a.jpg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
eafd062437bbb7c7319e2e30ecd7.jpeg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
c2f0cba689647125331c7f0d6e8f.jpeg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
cc7b3d0c29bda2a757052fe75828.jpeg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
345e977423e9b794f9b6caeba92a.jpg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
7dd553dc38e1ab3d3457497891bf.jpg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
03602ad3e5e8a28ff798b36d63fd.jpeg:1  Failed to load resource: the server responded with a status of 404 (Not Found)
dd4dbfe54487f188a46070267088.jpg:1  Failed to load resource: the server responded with a status of 404 (Not Found)

Hope this helps!

[ildyria]

that is the console. But can you check the network tab ? what is the url fetch there? :(

[jlgarnier]

Success! I made a fix in the compose file by adding the port to the APP_URL variable and it now works fine!
Sorry for the mess! I now have to get it working behind the proxy so I can access the app out my local network...

Thanks a lot for your help!

[ildyria]

Woop woop woop. 🥳

[ildyria]

Do you have a suggestion of something we should add on the docker_compose file for that ?

[jlgarnier]

I don't know, I'm not really familiar with the compose syntax, but it definitely needs a delay between the DB container running up and the app container creation...

[ildyria]

I can figure that one, that is easy. :)
I was more wondering about your port issue.

[jlgarnier]

Is there any way to add a file manually (file copy to the upload folder), then re-index all files of an album?
Regarding the port issue, my mistake: I juste replace the actual URL (pictures.mydomain.com) by the local IP address, forgetting about the port...

[ildyria]

Is there any way to add a file manually (file copy to the upload folder), then re-index all files of an album?

I assume you mean this: https://lycheeorg.dev/docs/faq_general.html#how-to-mass-import-photos ?

[ildyria]

Or this:

(I would recommend the docker exec way though).

[jlgarnier]

Oh, fine!
My photos are already organized in separate folders, named YYYY-MM-EVENTNAME. If I put them all in a /tmp/import folder, is there any way to automatically create an album from each subfolder (with matching name)? Or create the album with the --album_id option if it doesn't exist? Or add the corresponding tag to imported files, so I can easily sort and move them?

[ildyria]

If you select the /tmp/import folder, then all the sub folders will be created as sub albums etc, mirroring exactly the file structure you have in Lychee.

the --album-id is only if you want to import into that specific album instead of root.

[jlgarnier]

Seems you read my mind... 😁 OK, I'll try this and will let you know of the outcome!

[jlgarnier]

Got a strange message:

root@server5-VM-CasaOS:~# docker exec -it lychee-app php artisan lychee:sync /DATA/Gallery

Directories to sync:
  - /DATA/Gallery
Owner ID: 1
No album ID provided, importing to root album.
Import settings:
  - Delete imported files: No
  - Import via symlink: No
  - Skip duplicates: No
  - Resync metadata: Yes
Delete missing photos: No
Delete missing albums: No

Running in sync mode, this may take a while depending on the number of files.
Consider setting `QUEUE_CONNECTION=database` in your `.env` and using multiple queue workers (php artisan queue:work) for better performance.

Start tree-based syncing (maintains folder structure).
/DATA/Gallery: Building folder tree...
Given path is not a directory (/DATA/Gallery)

Actually, /DATA/Gallery IS a directory, which contains many subfolders... 🤔

[ildyria]

Is directory mounted in your docker compose O.o ?

[jlgarnier]

No, the source folder is on the host. How am I supposed to mount it? So far, there's only 3 volumes declared with the container:

• ./lychee/uploads:/app/public/uploads
• ./lychee/storage:/app/storage
• ./lychee/tmp:/app/storage/tmp

[ildyria]

You can mount as many folder you want. :)

- <path-to-data>/Gallery:/import-me

And then you can do:

docker exec -it lychee-app php artisan lychee:sync /import-me
(note that it will trigger an import-me album at the root)

[jlgarnier]

OK, thanks for the tip!