monerod excessive hard drive writes

[undeath]

I noticed that monerod writes a lot of data to my drives, a lot more than what would be expected from keeping the blockchain database up to date.

Within 120h (5d) of running my monerod IO stats (reported by systemd) are: IO: 204.7G read, 407.2G written

That's an average of almost 1mb/s. Those statistics only include keeping up to date with an already synced blockchain. This does not look right.

I'm running v0.17.3.0.

[moneromooo-monero]

Is this while syncing historical blocks or just keeping up with new blocks after initial sync ?

[undeath]

just keeping up with new blocks

[moneromooo-monero]

Thanks. It does seem excessive then.

[moneromooo-monero]

Here's something you can try, to see if it changes the write amount:

<+moneromooo> OK. Do you have a setting suggestion for --db-mode to see if that is the reason for the high usage ?
< hyc> I usually suggest fast:async:1000000
< hyc> but it also won't make much difference if there's not enough RAM to cache more dirty pages
< hyc> if they can run with /usr/bin/time -v that will give full stats from getrusage()
< hyc> you'd be looking for the pagefault counts to decrease with a looser sync-mode

So: --db-sync-mode fast:async:1000000

[undeath]

Looking much better with that setting. Node has been running for 22h now and current stats are IO: 33.9G read, 8.4G written. Also RAM consumption is lower (now 3.7gb, before 5gb).

[moneromooo-monero]

Still the same after a similar time from what you originally reported ?

[undeath]

Yes, significantly less writes with same amount of reads. After 137h (5.7d) the process is at IO: 212.7G read, 61.9G written. RAM consumption went up again and is now at 6.1gb.

[XfedeX]

Shouldn't the default DB config be changed?

[selsta]

@XfedeX then we get countless complaints again that the DB corrupts on Windows.. that's why it got changed in the first place.

[hyc]

Have considered changing the default back for everything except Windows...