Run s3_media_upload script

[nichu42]

Hey there,

I finally managed to have new media stored on S3 using the integrated S3 storage module.
Now I'm puzzled on how to move the existing media to S3.

I know this is done with the s3_media_upload script, which I found in /app/code/env/bin.

If I start the script with something as the suggested "s3_media_upload upload /app/data/data/media_store matrix --storage-class STANDARD_IA --delete", it comes back with the following error message:

s3_media_upload: error: Could not open 'cache.db' as sqlite DB: unable to open database file

I suspect that I am overlooking an important prerequisite. Anyone can help me out?

Best regards,

Nic

[robi]

First, find out where it's expecting the cache file. Then see if you can specify a location for it on the command line.

Alternatively, run it from the /app/data/ location in case it stores it right where it runs from.

[nichu42]

@robi
Thank you for your response.

I think I made a major step forward.

The script expects a database.yaml file to be present. It needs to include the user, password, database and host entries that can be found and copied over from the homeserver.yaml file.
Once you have prepared this, the script will create the cache.db on its own after you run the following command:

s3_media_upload update /app/data/data/media_store 1m

1m means all files that haven't been touched for a month will be used.

Now the script is ready to upload. This can be triggered with the following command:

s3_media_upload upload --delete /app/data/data/media_store s3_bucket_name

Unfortunately, here I am stuck again. The script returns the following error message:

botocore.exceptions.NoCredentialsError: Unable to locate credentials

The script documentation states "This module uses boto3, and so the credentials should be specified", and refers to https://boto3.amazonaws.com/v1/documentation/api/latest/guide/configuration.html#guide-configuration
Here I am completely lost again and don't even know where to start. Is this maybe something that needs to be done from Cloudron, @girish ?

[robi]

@nichu42 yeah those are s3 credentials for the bucket that need to be provided..

see if you can find example commands for the upload step.

[nichu42]

@robi
The example command is what I stated above.
The S3 credentials cannot be submitted on the command line.

They are in the homeserver.yaml, but it seems that boto3 (whatever that is) doesn't read them but expects environment variables to be set. That's why I thought this might be something that needs to be done by Cloudron.

[robi]

@nichu42 hmm, ok.

what's the error if you rename the .yaml file? (just to make sure that's where it's looking for the info..)

[nichu42]

@robi homeserver.yaml is the configuration file for Synapse. It will not start without it.
The S3 configuration is correct: Synapse uploads new media to the bucket.
However, "boto3" needs different configuration as it seems.

[robi]

@nichu42 right, here are the options:
https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html

[nichu42]

@robi said in Run s3_media_upload script:

@nichu42 right, here are the options:
https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html

Yes, that's what I figured.
But I have no idea how to make any of these options work with Cloudron.
The file system is read-only, so I cannot put a config file where "boto3" expects it (~/.aws/credentials).

That's why I thought that maybe @girish has to enable the use of environment variables.

[girish]

@nichu42 took me a while to figure what/where this script was. I guess it's this - https://github.com/matrix-org/synapse-s3-storage-provider/blob/main/scripts/s3_media_upload ?

[girish]

@nichu42 you have to create a so called database.yaml file manually as per https://github.com/matrix-org/synapse-s3-storage-provider#regular-cleanup-job

"database.yaml should contain the keys that would be passed to psycopg2 to connect to your database. They can be found in the contents of the database.args parameter in your homeserver.yaml."

From what I can make out from the code, it needs to be like this:

postgres:
    user: xx
    password: yy
    database: zz
    host: postgresql

Might be worthwhile asking upstream to document this...

[nichu42]

@girish Thank you for responding!

Yes, this thread is about the script that you have linked (https://github.com/matrix-org/synapse-s3-storage-provider#regular-cleanup-job). It is part of Cloudron's Synapse installation and can be found in /app/code/env/bin.

I had already managed to make the database config as you have mentioned in your post.

The problem is: The script uses "Boto3" (AWS SDK for Python) which expects the S3 credentials either to be saved in the config file ~/.aws/credentials or as environment variables, see
https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html

Please correct me if I'm wrong, but Cloudron doesn't grant me access to either of these. That's why I mentioned you in this thread. I think you'd have to enable one of these options to make the script work.

[girish]

@nichu42 From the link you posted, there is a bunch of environment variables you can set - both for the credentials itself and also for the config file. Have you tried those? Or is the question about how to use those env variables?

[nichu42]

@girish Yes, correct: How to set these environment variables with Cloudron?

[girish]

@nichu42 You are running this on a Web Terminal right ? You can just export foo=bar like in a normal terminal and then run the s3_media_upload script ?

[nichu42]

@girish Yay! Thank you.
I am all new to this Linux game so I wasn't aware I could just set the environment variables like that.

For everyone else, this is what you need to do:

──────────────────────────────

1. Set up S3 with Synapse. See my post here: https://forum.cloudron.io/post/60415

2. Create a database.yaml file in /app/data/configs that contains the postgres database credentials.
   You can find those in the existing homeserver.yaml file.

    user: xxx
    password: xxx
    database: xxx
    host: postgresql

3. Create a script (e.g., s3cleanup.sh) with the following contents:

#!/bin/bash
cd /app/data/configs
export AWS_ACCESS_KEY_ID=[your S3 compatible access key]
export AWS_SECRET_ACCESS_KEY=[your s3 compatible secret access key]
/app/code/env/bin/s3_media_upload update /app/data/data/media_store 1m
/app/code/env/bin/s3_media_upload upload --delete --endpoint-url https://yours3storageendpoint.com /app/data/data/media_store [your s3_bucket_name]

4. Run the s3cleanup.sh script.
   It will look up media that hasn't been touched for 1m (= 1 month) or whatever you set above. It needs to be an integeger value, followed by either m = month(s), d = day(s) or y = year(s).
   It will create a cache.db file that refers to the media that matches your criteria.
   In the second step, it will upload all files from the cache.db to your s3 storage and delete the local copies.

The output looks like this:

Syncing files that haven't been accessed since: 2022-12-25 14:59:14.674154
Synced 603 new rows
100%|████████████████████████████████████| 603/603 [00:00<00:00, 16121.24files/s]
Updated 0 as deleted
100%|████████████████████████████████████| 603/603 [03:25<00:00,  2.93files/s]
Uploaded 603 media out of 603
Uploaded 3203 files
Uploaded 263.6M
Deleted 603 media
Deleted 3203 files
Deleted 263.6M

Edit: Added path /app/data/configs to script to make it work as cron job.
Edit2: Added more choices for duration suffixes in 's3_media_upload update' job.

Disclaimer: This is to the best of my knowledge and understanding. It worked for me, but I accept no liability for loss of data on your server caused by my incompetence.