Workarounds required

[mmalgeri]

This is feedback:

1. Had to create a custom AMI and install scylla-enteprise-tools package to get cassandra-stress to work
2. Had to change the user from scyllaadb to ubuntu and make related changes to paths in variables.tf and scylladb-loaders.tf
3. Had to change line 42 in scylla-cloud.tf to peer_cidr_blocks = [var.custom_vpc]
4. Had to set fs.inotify.max_user_watches = 524288 and fs.aio-max-nr = 1048576 in /etc/sysctl.conf

After running the benchmark twice, the max ops/sec was around 400K, not 1M

Any other tweaking required?

[zseta]

Hi @mmalgeri

Regarding AMIs, depending on the region you use you might have to find a different ScyllaDB AMI (if the default one is not available in your region). But creating a custom AMI should not be required on your part.

I've merged a PR that fixes all the known issues. But seems like your issues came from having to create a custom AMI.

Normally, it takes a few minutes to reach 1M ops/sec