Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A beginner-friendly introduction to Prometheus

github.com

Post reply on HN
Print view View original

11–20 of 26 posts

Re: A beginner-friendly introduction to Prometheus

#11

Post by user5994461 »

therealmarv wrote: post #4

I wonder if it should be used at all for beginners and smaller teams. I thought it is something like the new industry standard but even to get it running in production is not trivial and had the feeling this is something for teams who are deeply committed to monitoring and developing their own metrics, views and alarms. My conclusion is the majority (or at least smaller teams) are maybe better served with something l…

Prometheus is just a storage engine, with a debugging page to display one metrics (it's what you'd get if you did a one week hackathon to make a metrics system really).

Datadog is a complete solution to collect metrics, visualize, dashboards, search and alerting. There is really no comparison possible (if datadog is a car, prometheus is an engine).

Prometheus is easy to setup but it barely does anything. It gets really messy when one needs to gather application metrics (postgresql, cassandra, haproxy, java, etc...) and have dashboards and alerts going to different places (slack, email, etc...). Trying to find and combine a variety of plugins and open source tools to fill the gaps.

Any company irrelevant of size should have a metrics/dashboard/monitoring/alerting solution. IMO datadog because it's better in all aspects. If the company doesn't want to pay or doesn't want SaaS, it should at least setup prometheus to have basic system metrics somewhere, rather than flying blind.

Re: A beginner-friendly introduction to Prometheus

#13

Post by collyw »

therealmarv wrote: post #4

I wonder if it should be used at all for beginners and smaller teams. I thought it is something like the new industry standard but even to get it running in production is not trivial and had the feeling this is something for teams who are deeply committed to monitoring and developing their own metrics, views and alarms. My conclusion is the majority (or at least smaller teams) are maybe better served with something l…

I tried to suggest something off the shelf to our place, as we have a lot of code and a small team. Our prima donna tech lead insisted that we need something custom for one small case (like everything) so we have Prometheus.

Re: A beginner-friendly introduction to Prometheus

#15

Post by throwaway50444 »

therealmarv wrote: post #4

I wonder if it should be used at all for beginners and smaller teams. I thought it is something like the new industry standard but even to get it running in production is not trivial and had the feeling this is something for teams who are deeply committed to monitoring and developing their own metrics, views and alarms. My conclusion is the majority (or at least smaller teams) are maybe better served with something l…

Prometheus scales down very poorly for small teams. Other monitoring systems are effective in few hour of work and require installing only one component.

Prometheus requires a plethora of tools around it to achieve feature parity and plenty of manual work to create dashboards, alerts and so on.

Not to mention the need for custom exporters for each application (!), contrasted to agent-based platforms that have built in plugins and autodetect what need to be monitored.

Re: A beginner-friendly introduction to Prometheus

#16

Post by dividedbyzero »

therealmarv wrote: post #4

I wonder if it should be used at all for beginners and smaller teams. I thought it is something like the new industry standard but even to get it running in production is not trivial and had the feeling this is something for teams who are deeply committed to monitoring and developing their own metrics, views and alarms. My conclusion is the majority (or at least smaller teams) are maybe better served with something l…

If you use GCP's Monitoring for dashboards and alerts on metrics from applications running in GKE, the easiest way to collect those seems to be running Prometheus with a sidecar container that ships Prometheus' metrics to StackDriver [1].

We're running such a setup currently, i.e. we installed Prometheus via Helm, using the Helm chart's sidecar config option to add the sidecar container, and the rest works more or less automagically (k8s workloads with Prometheus endpoints are detected based on annotations and their metrics show up in Monitoring)

We have somewhat relaxed availability constraints for the project this is part of, so we're using the opportunity to gain some experience with this setup and find out whether it's good enough for highly-available prod clusters. So far it's all nice and easy and works very well.

[1] https://cloud.google.com/monitoring/kubernetes-engine/promet...

Re: A beginner-friendly introduction to Prometheus

#17

Post by nopzor »

user5994461 wrote: post #11
therealmarv wrote: post #4

I wonder if it should be used at all for beginners and smaller teams. I thought it is something like the new industry standard but even to get it running in production is not trivial and had the feeling this is something for teams who are deeply committed to monitoring and developing their own metrics, views and alarms. My conclusion is the majority (or at least smaller teams) are maybe better served with something l…

Prometheus is just a storage engine, with a debugging page to display one metrics (it's what you'd get if you did a one week hackathon to make a metrics system really). Datadog is a complete solution to collect metrics, visualize, dashboards, search and alerting. There is really no comparison possible (if datadog is a car, prometheus is an engine). Prometheus is easy to setup but it barely does anything. It gets real…

prometheus is an amazingly versatile, performant, and high quality piece of software that does more than just storage (eg. service discovery, scraping, etc).

it’s tsdb engine (specifically the new one in 2.0) is pretty much state of the art.

i don’t think it’s fair to say that it’s the kind of thing that could be created in a one week hackathon.

Re: A beginner-friendly introduction to Prometheus

#19

Post by Hir0ki »

throwaway50444 wrote: post #15
therealmarv wrote: post #4

I wonder if it should be used at all for beginners and smaller teams. I thought it is something like the new industry standard but even to get it running in production is not trivial and had the feeling this is something for teams who are deeply committed to monitoring and developing their own metrics, views and alarms. My conclusion is the majority (or at least smaller teams) are maybe better served with something l…

Prometheus scales down very poorly for small teams. Other monitoring systems are effective in few hour of work and require installing only one component. Prometheus requires a plethora of tools around it to achieve feature parity and plenty of manual work to create dashboards, alerts and so on. Not to mention the need for custom exporters for each application (!), contrasted to agent-based platforms that have built i…

> Prometheus scales down very poorly for small teams

Cloud you maybe link some of those other tools, because my team is currently looking into monitoring software for our applications.

Post reply on HN