Backup mechanism sometimes fails to backup the database

[harivyasi]

I have observed this a couple of times on different machines. Problem seems to have appeared with the release of ELN version 1.10.x (1.10.5?).

Needs investigation.

[harivyasi]

After exploring one such system, the diagnosis seems to be that if the data volume is mounted on the filesystem (instead of as a volume) then it misses the file .version leading to the failure of the backup.sh command after compressing data, leading to skipping of the database backup step. The chemotion script still reports itself to be successful.

A fix can be to use the file from the /chemotion/app directory instead, as follows:

tar cvzH posix -f "/backup/backup-${stamp}.data.tar.gz" --directory=/chemotion/data/ . --directory=/chemotion/app .version

[harivyasi]

Another aspect is the modification of services.executor.volumes in docker-compose.cli.yml to match that of docker-compose.yml. This is because the CLI's backup command uses the executor container instead of the eln container.

[harivyasi]

A temporary workaround to fix this is to break the backup process in two parts:

• backup the database first: ./chemCLI instance backup --db, followed by
• backup of the data files: ./chemCLI instance backup --data.

Because the .version file is the last one to be backed up (and it is an optional file), the failure to find it does not impact backing up of other (necessary) files. And because the above command is a standalone command to backup the database, failure to find the .version file has no impact on the database's backup process.