Closed albertkohl-monotek closed 1 year ago
Change repository in your values
Example:
mailuVersion: 1.9.50
admin:
image:
repository: ghcr.io/mailu/admin
postfix:
image:
repository: ghcr.io/mailu/postfix
dovecot:
image:
repository: ghcr.io/mailu/dovecot
rspamd:
image:
repository: ghcr.io/mailu/rspamd
clamav:
image:
repository: ghcr.io/mailu/clamav
roundcube:
image:
repository: ghcr.io/mailu/roundcube
webdav:
image:
repository: ghcr.io/mailu/radicale
fetchmail:
image:
repository: ghcr.io/mailu/fetchmail
Release 1.0.0 has ghcr.io as default repository
That worked, thanks! my helm version was set to 1.9.10 and i was getting errors that it didnt exist, so i thought you didnt publish the v1 images based off what i saw in ghcr, but i might have been looking at that wrong.
if i wanted to bump to 2.0+ do i just do the same w/ the helm chart? if i'm not doing anything crazy in my helm chart, is it a simple upgrade? or any got-cha's i should keep an eye out for?
thanks!
Tried changing 'global.imageRegistry: ghcr.io', shouldn't that work rather than specifying each image path individually? (but it didn't)
Tried changing 'global.imageRegistry: ghcr.io', shouldn't that work rather than specifying each image path individually? (but it didn't)
It won't work as v0.3.3 of the chart didn't use global.imageRegistry
. FYI this won't work either if you want to change the registry for Mailu components in v1.x of the chart.
If you want to change the registry for the Mailu images in v1.x of the chart, use imageRegistry
(default is ghcr.io
)
It is so frustrating that the docker repos for 0.3.x have been pulled and could have been left for at least 6 months, that broke a working Kubernetes node when it tried to pull an image after a crashed pod. (went offline unexpectedly)
I then wanted to try and push on 0.3.3 chart with 1.9.50 as a quick fix to get the server back up (as my chart was 1.9.45), thats now been offline for 6 hours, and getting complaints to set roundhouse password (new one on me) and after that
Error: UPGRADE FAILED: template: mailu/templates/fetchmail/deployment.yaml:77:21: executing "mailu/templates/fetchmail/deployment.yaml" at <(default .Values.logLevel .Values.fetchmail.logLevel) "DEBUG">: can't give argument to non-function default .Values.logLevel .Values.fetchmail.logLevel
Command being attempted
microk8s helm3 upgrade --version 0.3.3 --install mailu mailu/mailu --values mailu-values.yaml -n mailu-mailserver --set global.database.roundcube.password=xyz
Yet nothing is set in fetchmail.logLevel in my chart.
It really was a move without thought about production machines. Why not put the docker images back for a while!!!
@jamesarbrown I understand your complaints about the images but that's not about the helm chart, it's about mailu. The best place to discuss this is the mailu matrix chat. Personally I understand both sides, mailu's (because docker inc made a change that cannot be answered other than dropping it) and yours (since your services are interrupted).
I have checked the chart source. There is no mailu/templates/fetchmail/deployment.yaml
in the 0.3.x charts, that a feature fro, 1.0.x. It seems that during the merge a build was triggered that released the initial 1.0.x release as 0.3.3.
I have re-released 0.3.3 from the correct branch as well as a 0.3.4 version (that is identical but allows an "upgrade"). Please try this one to get your service running again.
@fastlorenzo I made some manual pushes to gh-pages to get this sorted. Hope it does not break anything on the new builds.
Hi, I think i have just got it up.......
Then ran an adjusted Helm command to use the local files
microk8s helm3 upgrade mailu ./helm-charts/mailu --values mailu-values.yaml -n mailu-mailserver
This server is due to be migrated very very soon, hopefully some more detailed information as TBH no idea where Kubernetes stored the sqllite file.
Also as minor comment, on Chart 1.0.0 the single PVC is not being honoured for rspamd and redis, I am guessing these are now on Bitnami charts?
went into helm-charts/mailu/templates/fetchmail.yaml and took out the templating lines around line 65 (which was DEBUG and just left it DEBUG true etc)
Probably no need for this. The issue you had was because there was a completely different chart (1.0.x) in place of 0.3.3.
I just released 0.3.5 that points to 1.9.50 with the new image locations so that there's a more "official" workaround.
For reference: https://github.com/Mailu/Mailu/issues/2704
BTW: the primary reason to remove the old images was to force people to switch to the new repo as long as mailu has access to the docker-hub organization. Due to the change docker inc made, there was a high risk that the organization would be deleted after the grace period which could result in someone else getting control over the docker hub name "mailu" one day. In this case, instead of getting a broken deployment, you'd got an arbitrary crafted image from an unknown 3rd party on your server. So IMO deleting the images was the only choice for mailu and should be the choice for any project not willing to pay for docker organizations.
After docker inc received the shit storm the begged for, they reverted the change. But since they clearly showed their true intentions, migration away from that platform should be on anyone's roadmap.
Also as minor comment, on Chart 1.0.0 the single PVC is not being honoured for rspamd and redis, I am guessing these are now on Bitnami charts
For redis that's indeed the case, for rspamd I'll have to check why it doesn't because this is mailu specific and not using a subchart
Describe the bug image pull fails due t docker repo move
Environment microk8s
Additional context i noticed today that my mailu setup is not running. i used a helm chart to deploy 1.9.10. it appears that you guys moved the repo to github, but i dont see any images for 1.9.
[https://github.com/Mailu/Mailu/issues/2704](url)
is there any way to get those images posted as well so i can get mail up and runing?
secondary question, are their any plans on releasing a 2.0 helm chart?
Thanks!