ractf / polaris Goto Github PK

Native Image takes a million years to run and we install the JVM with the RPM anyway. Just use a shell script.

Container auto-updating has been AHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH because the Docker API is 💩. We need to design something better.

docker bad lol

The logs are spammed with the below and it makes them painful to use. Can we please exclude this endpoint from the logs...

[22/Apr/2021:07:55:56 +0000] "GET /healthcheck HTTP/1.1" 200 160 "-" "Consul Health Check" 3

I want to get rid of my folder of CURL requests to fix Polaris. Give me a real CLI please.

Every time /prometheus-metrics gets scraped, it triggers these warns. We should fix these, or failing that supress them with a filterfactory.

WARN  [2021-10-10 15:23:58,435] org.dhatim.dropwizard.prometheus.DropwizardMetricsExporter: Invalid type for Gauge jvm.attribute.name: java.lang.String
WARN  [2021-10-10 15:23:58,435] org.dhatim.dropwizard.prometheus.DropwizardMetricsExporter: Invalid type for Gauge jvm.attribute.vendor: java.lang.String
WARN  [2021-10-10 15:23:58,442] org.dhatim.dropwizard.prometheus.DropwizardMetricsExporter: Invalid type for Gauge jvm.threads.deadlocks: java.util.Collections$EmptySet

It would be nice if we could automatically scale down jobs which haven't been requested for X days. Trying to run 600 unused containers is not fun.

Check if /var/run/docker.sock is a symlink. If it is, the user is probably using Podman and we should WARN on that. I want to properly support Podman but we don't yet.

Probably because of namespacing