op-geth Archive Nodes Fall Behind Occasionally

[pc-quiknode]

It seems to us that op-geth archive nodes would fall behind full nodes by few hundreds blocks every few days. There's slight increase in CPU usage when the issue happens

Here's a pprof profile image taken for 30s of sampling when the issue was happening:

[wangjiangw]

It looks nobody will answer your question, I think the bottleneck is storage which can't be processed by the official team.

[ChanHongMing]

We are facing similiar issue, are there more specific hardware requirements on storage? We are already using NVMe as suggested in https://docs.base.org/chain/run-a-base-node

[ColeSu-n]

我遇到的问题时突然开始落后3000左右区块，而且本地节点区块高度还追不上,暂时不知道是什么原因

[wangjiangw]

我遇到的问题时突然开始落后3000左右区块，而且本地节点区块高度还追不上,暂时不知道是什么原因

@ColeSu-n 看来大家出问题的时间点都是一样的

[danyalprout]

If your Geth nodes are falling behind, the main things you can optimize on the hardware is low disk latency and single core CPU perf.

However we run i7i's for Geth and occasionally they still fall behind. I would recommend switching to Reth, you'll find it needs a lot less diskspace and has overall higher throughput.