Skip to content

Latest commit

 

History

History
99 lines (71 loc) · 3.44 KB

HOWTO.md

File metadata and controls

99 lines (71 loc) · 3.44 KB

How to setup backup-bench.sh yourself

The backup-bench script supposes you have a source system with a RHEL 8/9 clone installed (PRs for other systems are welcome). The default configuration will delete and create the following folders:

  • /opt/backup_test as the git dataset download folder
  • /backup-bench-repos as the folder which will contain backup repositories (local or on remote target)
  • /tmp/backup-bench-restore as the folder where backup restoration tests happen

You can customize those settings in backup-bench.conf file.

Local benchmarks

The script must prepare your machine by installing the requested backup software. You can do so with:

./backup-bench.sh --setup-source

It can run as local backup benchmark solution only, in that case you should run the following commands:

./backup-bench.sh --clear-repos
./backup-bench.sh --init-repos --git
./backup-bench.sh --benchmarks

The --git parameter for --init-repos command instructs the script to fetch the linux kernel source as backup source. This allows to have the same datasets for different workloads. You may also configure BACKUP_ROOT variable in backup-bench.conf to point to specific dataset and avoid using --git parameter.

You might want to run multiple iterations of backups. In that case, you can run the following

./backup-bench.sh --clear-repos
./backup-bench.sh --init-repos --git
./backup-bench.sh --benchmarks --git

Results can be found in /var/log/backup-bench.log and /var/log/backup-bench.results.csv.

Moreever, backup solution log files can be found in /var/log/backup-bench.[BACKUP SOLUTION].log

Remote benchmarks using SSH / SFTP backends

Remote benchmarks assume you have a second (target) machine. Both source and target machines must be reachable via SSH.

After having setup the necessary FQDN and ports in backup-bench.conf, you can initialize the target with:

./backup-bench.sh --setup-target

The target machine will connect to your source server to upload the ssh keys necessary for the source machine to connect to your target. This requires you to enter the password once. Once this is setup, the target cannot connect to source anymore.

Once this is done, the source machine can use the uploaded ssh keys to connect to the remote repositories on the target system. You can then prepare the benchmarks with

./backup-bench.sh --clear-repos --remote

Optional step if using kopia / restic HTTP servers, on target:

./backup-bench.sh --serve-http-targets

On source

./backup-bench.sh --init-repos --remote
./backup-bench.sh --benchmarks --remote

Again, you can run multiple backup iterations with:

./backup-bench.sh --clear-repos --remote

Optional step if using kopia / restic HTTP servers, on target:

./backup-bench.sh --serve-http-targets

./backup-bench.sh --init-repos --remote --git
./backup-bench.sh --benchmarks --remote --git

⚠️ duplicity benchmarks will fail if you do initialize repositories locally and try remote backup benchmarks or vice verca.

Remote benchmarks using alternative backends

There is a work in progress to support restic and kopia http servers, which have not been tested yet. You're welcome to help to automate those. Script will assume restic http, restic_beta http and kopia http ports are reachable from source to target. As of today, no auth mechanism is used in script, so please make sure you know what you're doing when using http backends.