Api does not work or works very slowly since 09/27

[wdsse]

Hi!
The API is down or very slow since 09/27.
For example, if I launch a node from the snapshot https://ops.store.eosnation.io/eos-snapshots-v8/snapshot-2024-09-27-01-eos-v8-0396348987.bin.zst, then the api will work for several hours and then start to slow down, and then stop working altogether. I didn't find any information in the logs. But if I launch from the snapshot https://ops.store.eosnation.io/eos-snapshots-v8/snapshot-2024-09-27-07-eos-v8-0396391839.bin.zst (or from another newer snapshot), then the api immediately stops working (or takes a very long time to respond).

Please tell me how to launch the node correctly to avoid such problems?

nodeos --full-version
info  2024-09-28T10:17:02.744 nodeos    main.cpp:158                  main                 ] nodeos started
v1.0.1-574650744460373f635d48cac9aa6dee67dcbfdb

lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 20.04.6 LTS
Release:	20.04
Codename:	focal

CPU - 6 cores
RAM - 20GB + SWAP 4GB
NET - 10Gbit/sec

[luoxiandong]

I have same problem,Has it been resolved？

[heifner]

This configuration is not powerful enough. Depending on your needs, the new recommendation for block producers is 192 GB RAM minimum with 256 GB RAM recommended. If running SHiP or trace API or a public relay node or public API node then increased CPU and disks may be needed.

[wangjiangw]

This configuration is not powerful enough. Depending on your needs, the new recommendation for block producers is 192 GB RAM minimum with 256 GB RAM recommended. If running SHiP or trace API or a public relay node or public API node then increased CPU and disks may be needed.

@heifner we also find this problem, but the cpu or ram are not fully used, the only diffrence we found is the IO counts are too high now.

[wangjiangw]

we are api node

[bhazzard]

As mentioned, this is because the state database on the EOS network has multipied in size due to usage. To successfully run nodes on the network requires more memory now. The new recommendation for is 192 GB RAM minimum with 256 GB RAM recommended.

This is needed even if you are not a block producer. But if you have sysadmin skills, some operators have had luck with lower memory configurations by leveraging tmpfs with a swap file.