It claims to be a distributed nosql kv store, very different to k8s. You might have just as much luck with eg. Clickhouse. Or boring old postgres.
I worked on some systems that supported Riak as a deployment option in 2013, but haven't since.
There isn't much recent news on it: https://aphyr.com/posts/285-jepsen-riak is from 2013 and Aphyr hasn't written about it in years. Armchair suspicion since we haven't heard about it in years is that it's pretty much end-of-life, deployed in a few legacy sites. But I'm curious too if it's still being considered for modern infra.
That was just the KV part, but it’s all built on top of Riak core and compute which does end up looking for like k8s. But yeah the reason I asked is because I stumbled upon it and wanted to know if it’s basically dead
Riak was one of the OG k/v stores by [basho](https://en.wikipedia.org/wiki/Basho_Technologies). They were kind of early to the world of distributed k/v stores and flourished quite a lot but ultimately got usurped by the cloud giants.
I was using Riak for IoT measurement storage. Was working well, apart from "not possible to delete something". We got rid of it, because it was WAY oversized for our needs, it requires at least 3 servers to get benefits of architecture and querying data was slow in practice and very unwieldy at start, because you needed to essentially write your own "sharding" and paging support for queries. It was essentially big dumping ground for data, that you might read less than write. We tried it at start because it sounded cool to us too.
We replaced it with postgres+timescale on one average server. Now after main contributing company folded, Riak is pretty much dead. But it never gave us any operational problems, was very stable.
I’m surprised it had a server minimum. Seems like it should architecturally scale from 1 to whatever. But I guess it isn’t as simple as the front page makes it seem. I assumed it works like a simple distributed compute and storage system
It's not real enforced minimum, only "minimum against server failure". You can run it on one server, but you won't get distributed failover storage (something like RAID).
Last I looked, Riak HTTP API let anyone execute arbitrary code, by design. Not an architecture I enjoy, even if they pretty much popularized vector clocks.
There is a core of maintainers who care about Riak (after the company Basho ended many years ago) and it appears that most work has been moved into the OpenRiak repository.
It claims to be a distributed nosql kv store, very different to k8s. You might have just as much luck with eg. Clickhouse. Or boring old postgres.
I worked on some systems that supported Riak as a deployment option in 2013, but haven't since.
There isn't much recent news on it: https://aphyr.com/posts/285-jepsen-riak is from 2013 and Aphyr hasn't written about it in years. Armchair suspicion since we haven't heard about it in years is that it's pretty much end-of-life, deployed in a few legacy sites. But I'm curious too if it's still being considered for modern infra.
That was just the KV part, but it’s all built on top of Riak core and compute which does end up looking for like k8s. But yeah the reason I asked is because I stumbled upon it and wanted to know if it’s basically dead
Riak was one of the OG k/v stores by [basho](https://en.wikipedia.org/wiki/Basho_Technologies). They were kind of early to the world of distributed k/v stores and flourished quite a lot but ultimately got usurped by the cloud giants.
Why do you think they lost out to cloud with such a head start?
I was using Riak for IoT measurement storage. Was working well, apart from "not possible to delete something". We got rid of it, because it was WAY oversized for our needs, it requires at least 3 servers to get benefits of architecture and querying data was slow in practice and very unwieldy at start, because you needed to essentially write your own "sharding" and paging support for queries. It was essentially big dumping ground for data, that you might read less than write. We tried it at start because it sounded cool to us too.
We replaced it with postgres+timescale on one average server. Now after main contributing company folded, Riak is pretty much dead. But it never gave us any operational problems, was very stable.
I’m surprised it had a server minimum. Seems like it should architecturally scale from 1 to whatever. But I guess it isn’t as simple as the front page makes it seem. I assumed it works like a simple distributed compute and storage system
It's not real enforced minimum, only "minimum against server failure". You can run it on one server, but you won't get distributed failover storage (something like RAID).
Do you have a link to what you're referring to? I'm only aware of Riak K/V database: https://en.m.wikipedia.org/wiki/Riak
Interesting website design: https://riak.com - reminds me of CSS Zen Garden. (Not that it says anything about how it performs etc.)
My experience with Riak is a NoSql datastore.
Last I looked, Riak HTTP API let anyone execute arbitrary code, by design. Not an architecture I enjoy, even if they pretty much popularized vector clocks.
https://aphyr.com/posts/224-do-not-expose-riak-directly-to-t...
Here is what the roadmap looks like at present.
There is a core of maintainers who care about Riak (after the company Basho ended many years ago) and it appears that most work has been moved into the OpenRiak repository.
https://github.com/orgs/OpenRiak/discussions/19
Unless strictly required for real for real, always use CP databases. Very very very rarely use AP dbs.
It was all the rage around 2012. Haven't heard of it since...