Closed alexttx closed 1 year ago
I think it'd be useful to have rollup stats to total bytes read/written by each job type so that we can easily see the distribution of media usage by job. Probably one per tree, with a rollup for the entire kvdb.
I think it'd be useful to have rollup stats to total bytes read/written by each job type so that we can easily see the distribution of media usage by job. Probably one per tree, with a rollup for the entire kvdb.
Persisted? I'm a little leery about persisting it. When would it be reset?
I guess the title is a bit misleading, but this RFC is focused on kvset stats. I didn't name it "kvset stats" since they are rolled up to node and tree level.
No, not persisted, just something we can dump via rest or at close. But if that's out of scope we can add it in whenever, we're really gonna need it when we have to starting figuring out what to throw out in the cloud environment (probably scatter remediation has to be disabled or set really high to let gc clean up scatter, or something....)
Some worthy ideas here, but this would be a large code change and our focus is on other areas of HSE.
Signed-off-by: Alex Tomlinson atomlinson@micron.com