Question: Why use loki-stack instead of separate promtail, loki and grafana installs?

[PhilipMay]

I want to install promtail, loki and grafana (and prometheus). Why should I use loki-stack instead of a separate promtail, loki and grafana install?

Is it because loki-stack makes sure that the versions are compatible to eachother? Is that right?

Maybe you want to add this as a "FAQ" to the loki-stack README.md so ppl. know why to use it.

[smerschjohann]

I'm stuck on the same question. loki-stack is quiet out of date, so I think this is no option at all. I disabled promtail and exchanged it with the latest version as it is able to support the cri / containerd kubernetes log format.

Now I discovered that loki is rather old as well. I try to figur out which setup is the best:

1. independent promtail, loki and grafana
2. promtail, loki-simple-scalable and grafana
3. distributed, canary?

What about the integration of loki + prometheus operator, are there any recommendations on that part?

[ChrisJBurns]

I'm also wondering the same. I use the stack and enabled prometheus and I get a warning in the UI stating: Using Loki as Prometheus data source is no longer supported and might stop working. Please use Loki data source for your Loki instance.. But all I did was use the helm chart. I didn't realise that the version of loki and promtail was old though. That was news to me. Perhaps the loki-stack chart needs updating? Or at least a a description of the difference between the helm charts and what the users would normally use them for.

[ChrisJBurns]

Update:

Since my last comment over an hour ago, I've recently found that installing loki with the loki-stack whilst disabling prometheus (default to false anyways) and grafana. So that the stack only installs loki and promtail, and then installing prometheus and grafana using the following more updated chart. https://artifacthub.io/packages/helm/prometheus-community/kube-prometheus-stack. I've found the error that I stated in my last comment now goes away.

This comment then shed some light on how to connect Loki up as a datasource in Grafana: https://github.com/prometheus-operator/kube-prometheus/issues/555#issuecomment-635938188

So the overal result ended up like

                                   grafana
                                      |
                |-------------------- | ----------------|
loki (data source)                              prometheus (data source)

Knowing this, I'll probably avoid using the loki-stack chart unless they update it to be more recent. So I'll just use the single loki chart instead.

[puppetninja]

I am struggling with outdated loki-stack also, but apparently it is still being maintained ? somehow loki-stack automatically configure the loki server for promtail for me, if I install them separately they are not working out of the box

[rufreakde]

What I see missing here are values.yaml defaults that work out of the box together for all helm charts! I would have loved to use:

• Grafana
• loki-simple-scaleable
• fluent-bit
• prometheus

But if you follow the official guide not even loki works correctly. https://github.com/grafana/helm-charts/issues?q=is%3Aissue+is%3Aopen+loki-simple-scalable

There are a lot of issues for this one helm chart alone.

loki-stack helm chart seems to at least work with default helm configuration and some simple persistance changes.

Maybe this is related to this (recommended) flag for this Tanka thingy... focus shift internaly?

[trevorwhitney]

FWIW, loki-simple-scalable is the most actively maintained by the Loki team at GL, and is the only helm chart we run in our own cloud (most of which is jsonnet, but a few clusters are using this chart). Could you elaborate on what problems you ran into setting up loki-simple-scalable? That chart is working for us currently, but as it's under rapid development, there's a chance the docs have drifted and we'd love help keeping them up to date!

[rufreakde]

@trevorwhitney

FWIW, loki-simple-scalable is the most actively maintained by the Loki team at GL, and is the only helm chart we run in our own cloud (most of which is jsonnet, but a few clusters are using this chart). Could you elaborate on what problems you ran into setting up loki-simple-scalable? That chart is working for us currently, but as it's under rapid development, there's a chance the docs have drifted and we'd love help keeping them up to date!

Hi sorry for my late reply but here is the issue with more details: https://github.com/grafana/helm-charts/issues/921

[trevorwhitney]

moved conversation to #921.