I wanted to do a test for backup and restore FDB from an existing cluster to a new cluster. I wanted to use the cycle load to test to do so.
- Run the cycle load on a new cluster.
- Backup the new cluster.
- Bringup another blank cluster
- Restore the backup on the blank cluster
- Run the cycle load on this cluster to validate that the restore was consistent to the backup.
Is my assumption about what the cycle load does correct? Is there any document on how to use the cycle load?
This is a clever use of the cycle test. Unfortunately, it wouldn’t work, as there’s no flag that you can currently pass to the cycle workload to ask it to only do the verification, and not the cycle generation and modification. Generating and modifying a cycle twice without clearing the database in between will look like a corrupted database to the cycle test.
I’ve posted a PR that would add a parameter that would let you do this, but that will only be in 6.2, so it doesn’t help your current 6.1.8 cluster.
Until then, you’d basically just need to do roughly the same thing by hand of load in some data while taking a backup, and verify that a consistent snapshot of it is there when you restore.
Thank you for getting back to me. I will follow up on your advise to do it by hand for now.
It turned out that you could just the validation step all along if you used
runSetup=false on cycle test, and I just didn’t know about it. It also happens that I failed to properly test my
checkOnly parameter, so it’s being reverted anyway. Sorry!
No problem. Thanks for letting me know