[BUG] Full node needs a large amount of IO for what it does

Describe the bug A clear and concise description of what the bug is.

Running a full node puts a relatively high random IO load on the disk with the .chia directory (on the order of hundreds of KiB/s) which is on an SSD in my case.
Having it on the HDD I’m plotting on actually causes the blockchain to fail syncing because it’s stuck waiting for IO.

My reasonably well informed guess is that it’s caused by blockchain db synchronisation.

To Reproduce Steps to reproduce the behavior:

1. Run a full Node
2. Run iostat 1
3. Observe burst of IO to main disk every few seconds

Expected behavior A clear and concise description of what you expected to happen.

A full node should have to sync to disk at most every few minutes; a loss of even a few minutes of blockchain state is negligible.

Screenshots If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: Linux
• OS Version/Flavor: NixOS 21.05
• CPU: Intel Celeron J4105

Additional context Add any other context about the problem here.

Running inside a Docker container with the chia dir in a Docker volume backed by ZFS.

[hoffmang9]

you get a block every 19 seconds and those have to be written out to both the blockchain db and the wallet db so that might now be “right.” That said, we’d love to squeeze some more performance out of here.

[mariano54]

This does seem like a lot, considering the DB is less than 100MB. There might be a bug somewhere