[Node Operator Question] Huge resource usage difference of op-geth for read only ethereum node and third party provider node

[aytunc-tunay]

Are you running the most up to date node software?

• Yes

Did you check the documentation?

• Yes

Did you check for duplicate questions?

• Yes

Issue Description

I'm running two different optimism node. The one on the left is connected to alchemy, the one on the right connected to our own besuteku read only ethereum node. Even though both of them have similar speed and block number (currently around 112 million) , their storage and ram usage so different, and I couldnt understand the reason behind. Can someone help me to understand this ?

Args for op-node connected to own read only ethereum node:

'--l1=http://besu-service:8545/'
'--l1.rpckind=basic'
'--l1.trustrpc'

Args for op-node connected to alchemy node:

'--l1=https://eth-mainnet.g.alchemy.com/'
'--l1.rpckind=alchemy'
'--l1.trustrpc'

Protocol Description

op-geth: us-docker.pkg.dev/oplabs-tools-artifacts/images/op-geth:v1.101304.1
op-node: us-docker.pkg.dev/oplabs-tools-artifacts/images/op-node:6a9f126e2f0503b55bd949fcdb7a3d7607ac0c9d
network: mainnet

Node Logs

No response

Additional Information

No response

Feedback

No response

[opfocus]

Just out of curiosity, are their sync modes the same? Are they all achieve mode or full mode?

[trianglesphere]

Bumping this @aytunc-tunay. If this issue is still around, could you get the head block hash & number from both of these op-geth nodes?

[aytunc-tunay]

Both are using this - '--syncmode=full'

[opfocus]

Is there any node set with --gcmode archive? Or the host of node running block explorer at the same time

[aytunc-tunay]

No, there is not.No, our Ethereum node is not set up in --gcmode archive. It is running Hyperledger Besu with --sync-mode=X_SNAP and --data-storage-format=BONSAI. We are not running a block explorer or any similar heavy services directly on this node

[opfocus]

Sorry, I don't know the reason. ....
I feel that the picture on the right is more reasonable.

[sbvegan]

I've asked around, initially we were thinking that because Alchemy might ve fetching all receipts individually so that might have been the cause of the data build up. But you're using the --l1.rpckind=alchemy so it should be optimized.

Can you increase the SSD size on the Alchemy node? It looks like its reaching capacity, maybe thats why the memory is working so hard? It looks like its reaching a threshold on those spikes and then the garbage collector is freeing up space.

I'm kind of stumped on this one.