Too many disk writes

[deemru]

Running node on the mainnet. For ~20 days got ~800 GiB of disk writes.

This doesn`t look rationally as a whole database just ~16 GiB in size. Raw blockchain in sqlite takes ~5 GiB.

leveldb-cache-size does not change anything.

Is it cause by checkpoints or what?

Can we tune disk writes somehow?

[phearnot]

Could you please tell me how you've measured the amount of data that's been written? We haven't noticed any excessive disk usage on our servers (running either testnet or mainnet nodes), but I'm willing to look into this.

[deemru]

• Windows 10 x64
• java version "1.8.0_162"
• java -Dwaves.directory=%BASEDIR%waves -server -Xms512m -Xmx4096m -jar waves.jar waves.conf
• run for few days then check java process writes

Today is 1 month of running node, check my stats (not all values look accurate, but i am checking disk SMART too, write values seem fair):

[phearnot]

Okay, so I can confirm this. Because of how LevelDB works, the amount of data written to disk exceeds blockchain (and state) size. There's a way to work around this, but it will take some time to implement. Thanks for your report!

[z642d]

There's a way to work around this, but it will take some time to implement.

Is there any solution for this issue? Can we config leveldb to reduce disk writes?
Node kills SSD drive so fast

[Karasiq]

@phearnot