Re:Keycloak – Open-Source Identity and Access Management Int...
by chromanoid on Monday June 19, 2023 @04:40AM
(#36388727)
Did you try the "new" version powered by Quarkus? I think it has a much smaller footprint.
[ Parent
| on HN ]
Re:Keycloak – Open-Source Identity and Access Management Int...
by RedlineTriad on Monday June 19, 2023 @05:24AM
(#36389009)
I tried the Quarkus version, but it does some kind of build step at container startup, and that is what used the most memory.
You can apparently build an "optimized" image[1] so it doesn't do it at runtime, but I didn't want to build a custom image, and I felt like the KeyCloak philosophy didn't align with my own.
In general I found myself avoiding Java applications because of their memory footprint.
[1] https://www.keycloak.org/server/containers#_creating_a_custo...
Re:Keycloak – Open-Source Identity and Access Management Int...
by brabel on Monday June 19, 2023 @07:21AM
(#36389744)
I understand your reluctance to use Java, but it's actually pretty good if the developers use small libraries instead of the giant frameworks most Java servers use... a Java server for this kind of stuff will run within 200MB+ comfortably and with very high performance when written properly. Anything using 3GB for this stuff is doing something terribly wrong.
Re:Keycloak – Open-Source Identity and Access Management Int...
by RedlineTriad on Monday June 19, 2023 @08:13AM
(#36390172)
Yeah I know, nothing except maybe the JIT that requires java programs to use more memory than Go, but in practice Java apps are almost always significantly worse.
C# is also pretty bad, but not quite as bad I think.
I can only assume it's a cultural thing since reflection and abstraction is used so much.
Rust apps are always nice since they are generally <20MB or something.
Re:Keycloak – Open-Source Identity and Access Management Int...
by dikei on Monday June 19, 2023 @07:21AM
(#36389741)
Probably not small enough for a home lab setup, between a 50MB Golang App and 500MB Java App I'd choose the Golang App every time for my home lab, because memory is precious and probably less than 10 people will use it.
However, it's a different story when picking the solution for my day job, where RAM is abundant, but developer time is in short supply.