V1.4.0 Failed to parse /etc/grafana/grafana.ini, open /etc/grafana/grafana.ini: no such file or directory

[Robbilie]

Failed to parse /etc/grafana/grafana.ini, open /etc/grafana/grafana.ini: no such file or directory

there is no such file in there since the config dir is in /usr/share/grafana
theres at least a defaults file…

[andyxning]

@Robbilie there is a PR will add a default and empty grafana.ini if not exist already. See this diff for more info.

BTW, you can mount a configuration file into /etc/grafana/grafana.ini at runtime.

[Robbilie]

i know that i can mount a config file into it but i would rather not.
i have no idea why the config file is missing and this was really annoying when i tried to upgrade to 1.4.

any eta when a release with the fix is expected?

[andyxning]

@Robbilie The fix can only add an empty grafana.ini which may not be what you want. The desired way is to mount a customized grafana.ini. :)

[Robbilie]

thats not the desired way. the official grafana and previous versions of heapsters grafana had a grafana.ini and i dont see why the 1.4.0 release shouldnt have one

[andyxning]

@Robbilie Having not know that we have a grafana.ini in previous release. Will check this later.

[Robbilie]

lol theres literally no reason to not have one, 1.3 aswell as official grafana releases had one

[andyxning]

@Robbilie You mean gcr.io/google_containers//heapster-grafana-amd64:v4.0.2

[Robbilie]

@andyxning with one less slash yeah, thats the previous one which had the ini, https://hub.docker.com/r/grafana/grafana/ has the ini aswell

[trunet]

any workaround? trying to deploy a brand new heapster and having problems

[Robbilie]

mount the file into the pod…
although this should imo be fixed ASAP in a 4.2.1 release bc this is just annoying and breaking…

[DirectXMan12]

I believe @piosz is working on it, and it's marked as high priority. We should have a fix soon.

[loburm]

/assign @loburm

[doubaokun]

Change the image to be gcr.io/google_containers/heapster-grafana-amd64:v4.0.2 will fix this issue.

[Robbilie]

thats not a fix, i used to use a newer grafana image while waiting for heapster 1.4 because i need certain grafana notification channels

[piosz]

Apologies for the problems. @loburm is investigating this.

[trunet]

any news?

[loburm]

Sorry for the delay, I'm looking at this problem now. I will post update as soon as will find a root cause of issue.

[loburm]

This commit on the side of grafana is the root cause of the issue:
grafana/grafana@b191638

Shortly: previously they have been always creating grafana.ini file but in January this part was moved to the postinst script which is executed only when we actually install package on the system (in our case we just extracting files, so script is not executed).

I'll try to find workaround. Maybe we can use official grafana image for this.

[loburm]

Unfortunately, solution with docker image officially provided by the grafana is not going to work. This image is built only for amd64 (we support different arch) and image size almost 2 times bigger.

So I have prepared PR that just manually copies configuration files to /etc/grafana/: #1728

[ffahri]

check this #1722
v4.0.2
change to --> image: gcr.io/google_containers/heapster-grafana-amd64:v4.0.2

[loburm]

@ffahri this is not a fix of problem. Some users need features that have been released in 4.2.0 or later we just can't stuck with version 4.0.2 forever.

[loburm]

PR was merged. Tomorrow I'll push a new image to the gcr.

[loburm]

Image is pushed. In #1731 updates deploy file.

[discostur]

@loburm just tested the new image and it works without any problems ;)

[loburm]

@discostur thanks for checking it.

Considering that last PR was merged, I'm closing this issue.

[emansom]

It works again. Thanks guys. 👍

[jpds]

As a heads-up, this is still broken in the https://github.com/kubernetes/heapster/tree/release-1.4 branch

[andyxning]

@DirectXMan12 @piosz @loburm we should cherrypick #1728 to release-1.4.

@jpds As a workaround, give v1.4.1 a try.

[BruceShenC]

@andyxning I have tried but it still fail