## S3 Storage
PF_ENABLE_CLOUD=true
FILESYSTEM_DRIVER=local
FILESYSTEM_CLOUD=s3
PF_LOCAL_AVATAR_TO_CLOUD=true
MEDIA_DELETE_LOCAL_AFTER_CLOUD=true
AWS_ACCESS_KEY_ID=xxx
AWS_SECRET_ACCESS_KEY=xxx
AWS_DEFAULT_REGION=xxx
AWS_BUCKET=mediapixelfedsg
AWS_URL=https://media.pixelfed.sg
AWS_ENDPOINT=https://xxx.backblazeb2.com
#AWS_USE_PATH_STYLE_ENDPOINT=false

and it works.

Anyone else able to figure anything else out with the profile avatar and using S3 compatible storage?

If I'm understanding this correctly (and I could be wrong), this issue was addressed by this commit back in July 2022. And was bundled into the v0.11.4 release, which was made available in Oct 2022.

Wonder if it'll auto switch to s3 without touching.
Since it's storing images on both locations, will it like auto delete from server?

I think it's supposed to do that after 12 hours or something , but I've not checked properly to see if that is indeed the case .

then when you run php artisan cache:clear it'll get fixed.

Wonder if it'll auto switch to s3 without touching.

Since it's storing images on both locations, will it like auto delete from server?

@jdaviescoates I'm stuck on the URL still being munged with half local and half s3 with the profile pic...

Ah, I think removing/ commenting out AWS_URL might fix that.

Can't remember exactly but I had it and then later after doing something, I didn't.

I think perhaps it was that, or it that plus changing FILESYSTEM_DRIVER from s3 to local.

@jdaviescoates yep - it's a bit of an odd situation. and blocking me from going live

I'm not expecting to be swamped with users and so I've decided to go live with having everything store twice for now because I've a fair bit of unused disk space right now. If I end up getting a load of active users then I may need to manually start deleting stuff from local storage until this is fixed, but hopefully it'll be fixed before then!

Hope they get that bug fixed

Like it to help them pay attention to is?

The obvious solution would seem to me to be to simply make avatars/ profile images work when FILESYSTEM_DRIVER=s3 too, then all would be fine and dandy imho.

presumably eventually that'll get deleted, ah but then again perhaps not as per https://github.com/pixelfed/pixelfed/issues/3845#issuecomment-1331691661 ).

I've filed an issue about this:

https://github.com/pixelfed/pixelfed/issues/4114

I've also confirmed that manually deleting images from the local storage via the File Manager does not break posts.

So I don't know if this is a bug or if I've got something wrong in my config?

Reading this https://github.com/pixelfed/pixelfed/issues/4041

It sounds like you have to have FILESYSTEM_DRIVE=local for the Avatars to work. So I tried that.

But now it seems all files are getting stored twice, once locally and then again in the Object Storage.

See e.g. https://bridport.pics/i/web/post/522792891553329155

Really I only want to store thing in the object storage, not locally at all. But I've got MEDIA_DELETE_LOCAL_AFTER_CLOUD=true so maybe eventually it will delete the local versions?

Yep, it seems that perhaps files are just being cached locally, because if I open the Web Terminal and run php artisan cache:clear then right-click Open image in new tab it opens the S3 version (although as you can see in the post linked to above, for now the local version still seems to exist too... presumably eventually that'll get deleted, ah but then again perhaps not as per https://github.com/pixelfed/pixelfed/issues/3845#issuecomment-1331691661 ).

https://bridport.picshttps//bridport-pics.s3.fr-par.scw.cloud/public/avatars/052/274/889/869/491/404/9/3KE8LIMTBRbT8dlgbysN_avatar.png?v=3

If I take of off the superflous https://bridport.pics at the beginning it works fine:

https://bridport-pics.s3.fr-par.scw.cloud/public/avatars/052/274/889/869/491/404/9/3KE8LIMTBRbT8dlgbysN_avatar.png?v=3

So I wonder how/ why that extra bit is coming from...

I can see this in the logs:

Jan 22 13:59:19 [2023-01-22 13:59:18] production.ERROR: Image source not readable {"exception":"[object] (Intervention\\Image\\Exception\\NotReadableException(code: 0): Image source not readable at /app/code/vendor/intervention/image/src/Intervention/Image/AbstractDecoder.php:351)

So I don't know if this is a bug or if I've got something wrong in my config?

Once upstream makes a new release with the fix, we will update the package of course.

I'm confused though, because that fix was committed back in July 2022 and is tagged Pixelfed 0.11.4 which is also what we're running on Cloudron, so I'm unclear why Profile pics with S3 enabled aren't working?

but when it comes to profile photo change. it doesn't work.

Ah, but I'm also hitting this issue.

Did you manage to get it solved @ehsan0921 ?

But it's not working.

It is now!

I looked my settings for Mastodon for some clues and noticed that in there the endpoint included https://

All I had to do was to add that and then it started working! I wish there were most established standards around this. Some apps seem to want/ need the https:// in the endpoint and others don't so it become a matter of trail and error.

This is my working config in env.production:

# Store media on Scaleway S3 object storage
PF_ENABLE_CLOUD=true
FILESYSTEM_DRIVER=s3 
FILESYSTEM_CLOUD=s3
PF_LOCAL_AVATAR_TO_CLOUD=true
MEDIA_DELETE_LOCAL_AFTER_CLOUD=true
AWS_ACCESS_KEY_ID=xxx
AWS_SECRET_ACCESS_KEY=xxx
AWS_DEFAULT_REGION=fr-par
AWS_BUCKET=bucket-name
AWS_URL=https://bucket-name.s3.fr-par.scw.cloud
AWS_ENDPOINT=https://s3.fr-par.scw.cloud
AWS_USE_PATH_STYLE_ENDPOINT=false

I raise this issue in GitHub:

Hi guys

I linked S3 as below.

PF_ENABLE_CLOUD=true
FILESYSTEM_DRIVER=s3
FILESYSTEM_CLOUD=s3
AWS_ACCESS_KEY_ID=******
AWS_SECRET_ACCESS_KEY=*******
AWS_DEFAULT_REGION=ap-southeast-2
AWS_BUCKET=********

everything is fine with post. all uploading on S3 with no issue.
but when it comes to profile photo change. it doesn't work.
seems the address have some issue.

Hi @ehsan0921 (or anyone else who has this working) do you think you could share more about your Pixelfed S3 set-up?

I'm trying to get it set-up with my Pixelfed install too but struggling to get it to work.

You don't appear to have specified any AWS_URL nor AWS_ENDPOINT but I'm presuming you must have at least one of those in your config too, right?

I've currently got this in my env.production

# Trying to store media on Scaleway S3 object storage
PF_ENABLE_CLOUD=true
FILESYSTEM_DRIVER=s3 
FILESYSTEM_CLOUD=s3
PF_LOCAL_AVATAR_TO_CLOUD=true
MEDIA_DELETE_LOCAL_AFTER_CLOUD=true
AWS_ACCESS_KEY_ID=xxx
AWS_SECRET_ACCESS_KEY=xxx
AWS_DEFAULT_REGION=fr-par
AWS_BUCKET=bucket-name
AWS_URL=https://bucket-name.s3.fr-par.scw.cloud
AWS_ENDPOINT=s3.fr-par.scw.cloud
AWS_USE_PATH_STYLE_ENDPOINT=false

But it's not working.

I see in the logs:

Jan 22 13:18:46 [previous exception] [object] (GuzzleHttp\\Exception\\ConnectException(code: 0): cURL error 6: Could not resolve host: bridport-pics. (see https://curl.haxx.se/libcurl/c/libcurl-errors.html) at /app/code/vendor/guzzlehttp/guzzle/src/Handler/CurlFactory.php:200)

@scooke have you done this with Pixelfed too?

Once upstream makes a new release with the fix, we will update the package of course.