Prototyping is cheaper than ever, so I tried to imagine the simplest possible ClickHouse service.
The result is a side project called ClickHamlet. You run:
clickhouse client --host https://t.co/lfu0I1aDkR --secure --user <name> --password
And that's it. No sign-up form, real ClickHouse in the cloud.
me estoy metiendo en la cama y me acabo de acordar de que me he dejado la bici atada en la calle en plena Barcelona
mañana sigue entera? dadme esperanzas
clickhouse's http interface now works on @ClickHamlet, proxied straight to the backends. anything that talks http to clickhouse just works: curl, clients, agents
to prove it, https://t.co/49d5KN93Py runs on it
@marchoeffl makes sense, owning the engine gives you a lot more room. i stayed on stock clickhouse + plain_rewritable, regular servers from the official image. is the nvme cache per node or shared across readers?
i wrote about building a serverless clickhouse. every table lives on S3 with plain_rewritable disks, servers start on the first query and stop when idle, and there's no sign-up: one clickhouse client command and you have a database. how it works 🧵
hackernew is so random. somebody published the post in here: https://t.co/4dZvf8DkRw
when i saw it (due to a bunch of visits to the website) i wrote a comment with the usual "author here, happy to aswer any questions" and got flagged and removed
backends run on spot, writers included. they hold no data, so a reclaimed node looks like a crash: running queries fail, the next one starts the backend on another node, and everything acknowledged is already in S3