On docker volume restore: could not access file "$libdir/timescaledb-1.6.0": No such file or directory

Question

I am trying to restore TimescaleDB data from one ubuntu machine to another ubuntu machine. On the both machines we have deployed TimescaleDB through docker and using the same tag.

Docker image used: timescale/timescaledb:latest-pg11

I am creating a tar file for backup for the docker volume on the old machine, using below command:

docker run --rm -v $VOLUME_NAME:/volume -v /tmp:/backup busybox tar cvf /backup/$VOLUME_NAME.tar -C /volume ./

Also I am restoring using below command on the new machine:

docker run --rm -v $VOLUME_NAME:/volume -v $(pwd):/backup busybox sh -c "rm -rf /volume/* ; tar -C /volume/ -xvf /backup/$VOLUME_BACKUP_FILE"

Database restore was successful but when I see the docker logs I am getting below error:

2021-03-06 06:28:55.064 UTC [239] ERROR:  could not access file "$libdir/timescaledb-1.6.0": No such file or directory
2021-03-06 06:28:55.064 UTC [239] STATEMENT:  CREATE TABLE "__migration_lock" ("id" int NOT NULL, CONSTRAINT "PK_1312144c987e032cce38cfa00a2" PRIMARY KEY ("id"))

Do you know why this error is coming and how to resolve it?

Thanks

Answer 1

Docker images tagged with latest contain the latest release at the time you retrieve it (this is Docker's convention), thus the image in the source contains different TimescaleDB's version (1.6.0) than in the current latest image (2.1.0 is latest now). This leads into the error of missing binary, since the images are not the same and the current latest image doesn't contain 1.6.0, which was latest some time ago.

The easiest solution is to download the image with the same TimescaleDB version as in the source. I believe timescale/timescaledb:1.6.0-pg11 should work in your case.

You can double check, which TimescaleDB's version is in use, by calling \dx in each database.

Note that a TimescaleDB's image usually contains 5 older versions before the tagged one. So if you use image tagged 1.7.3-pg11 it is likely to work too, since it expected to contain binaries for 1.6.0-pg11. This is important to know if you are going to update TimescaleDB version from 1.6.0 to later, since you will need to use several images to get to the current latest version.

In general to avoid confusion it is better to use an image with specific version tag.

Answer 2

After you've done the restoration on the same version of TimescaleDB as the one on which you created the backup, here's how you can update your TimescaleDB extension in Docker.

First ensure you've got the latest Docker image installed:

docker pull timescale/timescaledb:latest-pg11

Start the container as you've done before.

Connect to the running Docker container and start psql immediately:

docker exec -it <container> psql -X -U postgres

Connect to the database you want to update (e.g. for a DB called mydb) to update its TimescaleDB extension:

postgres=# \c mydb
You are now connected to database "mydb" as user "postgres".

Update the extension:

mydb=# ALTER EXTENSION timescaledb UPDATE;
ALTER EXTENSION

Check the version now!

mydb=# SELECT default_version, installed_version FROM pg_available_extensions where name = 'timescaledb';
 default_version | installed_version
-----------------+-------------------
 2.3.0           | 2.3.0
(1 row)