mirror of
https://github.com/SebastianWendel/nixpkgs.git
synced 2024-09-21 12:59:04 +02:00
175cc7efd2
I read this hilarious blog post: https://wejick.wordpress.com/2022/01/29/can-i-have-a-smaller-prometheus/ We can have a smaller Prometheus too. This patch allows users to remove service discovery for five public clouds (AWS, Azure, DigitalOcean, GCP, and Linode) and also Kubernetes, simply by setting the corresponding enable-flag to `false`. I have tested building with each flag as I added it to the list. I also tested running with all six flags set to `false`, and the resulting Prometheus can still handle my orthogonal service-discovery configuration (files). To meet Adam Savage's definition of science, I measured the size of the `prometheus` and `promtool` binaries after adding each flag with `ls -h`. flag | prometheus | promtool --------------|------------|---------- starting size | 84M | 74M AWS | 72M | 61M Azure | 71M | 61M GCE | 64M | 53M k8s | 40M | 53M DO | 39M | 52M Linode | 38M | 51M I did not go as far as the blog post. If folks want, I'll make the rest of the service discovery optional too. I did not shrink the build closure, just the output closure; we still pull all of the various vendored modules into the Nix store during builds. I don't see how to do this in a neat or easy way. |
||
---|---|---|
.. | ||
alerta | ||
alertmanager-bot | ||
alertmanager-irc-relay | ||
bosun | ||
buildkite-agent-metrics | ||
cadvisor | ||
consul-alerts | ||
do-agent | ||
fusion-inventory | ||
grafana | ||
grafana-agent | ||
grafana-image-renderer | ||
grafana-reporter | ||
heapster | ||
icinga2 | ||
kapacitor | ||
lcdproc | ||
loki | ||
longview | ||
mackerel-agent | ||
matrix-alertmanager | ||
mtail | ||
munin | ||
nagios | ||
net-snmp | ||
newrelic-sysmond | ||
plugins | ||
prometheus | ||
riemann | ||
riemann-dash | ||
sensu | ||
sensu-go | ||
seyren | ||
telegraf | ||
thanos | ||
uchiwa | ||
unifi-poller | ||
zabbix | ||
zipkin |