Restoring Mnesia Database

[Arokan13]

Hey,

I'm using the docker-container on arm64, where before I used an old container v15.04.
I dumped the mnesia db to binary and made a backup in parallel and I'd like to restore the data.
Now, in order to do that, the server must be stopped which in turn stops the container, so I can't use ejabberdctl anymore.
Do you guys have an idea how to resolve this issue?

Also is there a way to export the mnesia db to another database as mnesia is limited to 2GB?

Thanks in advance!

[badlop]

I can think several ways to backup your mnesia database and later restore it in another machine:

A) Stop ejabberd, copy the mnesia spool files to the new location, then start the new ejabberd.

B) Use the backup and restore or install commands, which work with binary files. They require ejabberd be running.

C) Use the dump and load commands, which work with text files, only suitable for small servers. Require ejabberd running

D) Use export_piefxis and import_piefxis, which work with XML files, this may miss information as it supports only that XEP.

E) Use export2sql command, and later import in an SQL database.

I dumped the mnesia db to binary and made a backup in parallel and I'd like to restore the data.

Great!

Now, in order to do that, the server must be stopped

No. If you used the backup, dump or export_piefxis commands, you are expected to start the new ejabberd node, and then run the corresponding command: restore/install, load, or import_piefxis

Also is there a way to export the mnesia db to another database as mnesia is limited to 2GB?

Option E)

[jdolan-chwy]

Hi,

I have a similar question. I'm using an ejabberd Docker container from within unit tests which create various users, MUC rooms, etc. I'm hoping to take a backup of the database when the container first comes up and my admin account is created, and then restore that backup after each unit test runs. This will ensure that each test has a clean slate ejabberd instance to work with. However, this is not working as I'd assumed it would:

17:30:20.890 - DEBUG [main] EjabberdSupport : ejabberdctl backup /tmp/db -> Container.ExecResult(exitCode=0, stdout=, stderr=)

[ .. test runs here ]

17:30:46.702 - DEBUG [main] EjabberdSupport : ejabberdctl restore /tmp/db -> Container.ExecResult(exitCode=0, stdout=, stderr=)
17:30:47.595 - DEBUG [main] EjabberdSupport : ejabberdctl clear_cache -> Container.ExecResult(exitCode=0, stdout=, stderr=)

No errors, but on the very next test case that runs, it appears as tho the users and MUC rooms from the previous test are still there. Is there a way to purge ejabberd of data without restarting the container between each test case?

[jdolan-chwy]

Follow-up, I've tried both backup / restore and dump / load. Basically the same behavior with both sets of commands.

[badlop]

No errors, but on the very next test case that runs, it appears as tho the users and MUC rooms from the previous test are still there.

The ejabberd function that implements the restore API command explicitly keeps existing content, using the keep_tables option when calling mnesia.

Is there a way to purge ejabberd of data without restarting the container between each test case?

Yes. For cases like yours, you may want to use delete_mnesia before restoring a backup.