I have a question about how fdbbackup and fdbrestore work. My source db has 12 storage node and is running a 3 data hall setup. Is it possible to run a fdbrestore on a backup data created on that source db to a different setup of the database like a new fdb cluster with only 4 storage nodes and running in single/double config?
Yes, it works out of box.
Yes, it works but I will need to increase the memory of the destination cluster, otherwise it will hang.
Do you know is there a formula on how much memory is require vs the backup size?
My backup has about 40GB of data which roughly translate to 68537 blocks
Another interesting thing I see is that the restore will reach to a certain point in BlockInProgress like 28841, then it will go back down but the running block count and the byte written will still increase. So, what’s that BlocksInProgress actually means? The cached data in memory before writing back to the database?
The standard recommendation (and default --memory setting) for all FDB clusters is 8GB per process. There’s no scaling of memory with backup size that should be needed.
If I use the default, the restore hang in the middle. After I increase the memory to 12G, it no longer hang.
The backup is about 50GB in size.
What storage engine are you using on the target database?
The storage engine is ssd-2