3 ms·
Why is the pull model bollocks? I’ve been building monitoring for stuff since nagios and zabbix were the new hot tools; and I can’t really imagine preferring th
by cyberpunk 1y ago
Why is the pull model bollocks? I’ve been building monitoring for stuff since nagios and zabbix were the new hot tools; and I can’t really imagine preferring the oldschool ways vs the pretty much industry standard of promstack these days…
- KaiserPro 1y agoZabbix is bollocks. so is nagios. Having remote root access to all your stuff is utter shite. Prometheus as a time series DB is great, I even like its QL. What I don't like is pull. Sure there is agent mode or telegraf/grafana agent. But the idea that I need to hold my state and wait for Prometheus to collect it is utterly stupid. The biggest annoyance is that I need to have a webserver somewhere, with a single god instance(s) that can reach out and touch it. Great if you have just one network, but a bollock ache if you have any kind of network isolation. This means that we are using influxdb and its shitty flux QL (I know we could upgrade, but thats hard)
- cyberpunk 1y agoeh in a standard three tier you’re usually okay to pull up and push down aren’t you? Run it in the lower network.. Were all kubernetes these days so i guess i didn’t think about it a lot in recent years.
- nullify88 1y agoAFAIK, you can now send OpenTelemetry directly to Prometheus. So effectively it supports both push and pull models.
- dengolius 1y agoAFAIK, OpenTelemetry has never been effective compared to Prometheus.
- mystifyingpoi 1y agoBoth models are totally fine, for their specific use cases.