Fast catchup not synchronizing

post by liva on Jun 15, 2021

Hello, I am facing a problem when trying to install a node.

The fast catchup process gets stuck at 4823552 processed accounts for a while then restart at 0:

This is what I currently have:

Last committed block: 1092
Sync Time: 66622.4s
Catchpoint: 14410000#Q7CJTIBXVIF2PAT2C2546VJXGI26ARJE6N3YE3BHY7FCXIBIXSQA
Catchpoint total accounts: 11659041
Catchpoint accounts processed: 4823552
Catchpoint accounts verified: 0
Genesis ID: mainnet-v1.0
Genesis hash: wGHE2Pwdvd7S12BL5FaOP20EGYesN73ktiC1qzkkit8=

The version output:

liva@algonode:~# algod -v
8590327808
2.6.0.stable [rel/stable] (commit #8fe22d4e)
go-algorand is licensed with AGPLv3.0
source code available at https://github.com/algorand/go-algorand

My IO activity is saturated, algorand is using 99% of it as I am not on an SSD drive.

The CPU and the RAM are just okay. Should I assume my hard drive is just too slow?

Tho, I already installed a node on this same system like 3 or 4 months ago without any issue.

I’m currently searching into the logs for some more information.

Any help would be appreciated, thanks in advance!

post by tsachi on Jun 15, 2021

Hi @liva,

During the processing of the accounts, the primary function of the node is to decompress the incoming catchpoint file data stream and expand the individual accounts to disk.

As you could imagine, this has two common bottlenecks: the disk I/O throughput and the network bandwidth. Given that you already know that the machine you’re using doesn’t meet the recommended specification, I would suggest you try to repeat that with a fast SSD drive.

Note that there are SSD drives that are too slow. (Generally speaking, 5-6 years old SSD drives were pretty slow compared to modern ones)

If you were to buy an SSD drive for that purpose, I would recommend the Samsung 970/980 Pro NVME. It provides a good cost/performance ratio.

post by liva on Jun 15, 2021

Thanks a lot for this detailed answer.

I will go check for this SSD