We’re running Mender 3.7.7 (Open Source) on EKS using Helm chart v5.11.2. Everything was working fine until a recent cluster migration — our MongoDB and Redis pods stopped pulling because they rely on Bitnami images, which Broadcom is deprecating and moving behind a paywall.
We were able to pinned the image digests that were still cached on the nodes as a temporary fix, but we know this won’t last.
What would be the recommended path to overcome this situation?
A few specific questions:
Has anyone successfully migrated their MongoDB data from the Bitnami sub-chart to an external MongoDB instance?
We’re on the Open Source version but Redis is still running — is it actually needed, or can we safely disable it?
Any gotchas we should be aware of?
Our current setup:
Mender: 3.7.7 (Open Source)
Helm chart: v5.11.2
MongoDB: Bitnami sub-chart
Redis: Bitnami sub-chart
Any advice or shared experience would be really appreciated.
Thanks!
Has anyone successfully migrated their MongoDB data from the Bitnami sub-chart to an external MongoDB instance?
This mostly involves specific MongoDB competencies. For just Mender Server, probably it’s better to run a fresh new install with the latest helm chart and then move the functionalities and the devices to it and then decommission the old Mender Server 3.7.7.
We’re on the Open Source version but Redis is still running — is it actually needed, or can we safely disable it?
If you use the last helm chart (v7.x and ongoing) you can safely let it disabled for Open Source version (default behavior)
Any gotchas we should be aware of?
To address the bitnami change, the Mender Server Helm Chart v7.0.0 includes sample MongoDB and Redis manifest for convenience; please note that you should use them only for evaluation purposes, as there are no HA features like the Bitnami charts provided
Verify our devices reconnect without re-provisioning
Before we go ahead, a few things we’d love to confirm:
Can we jump directly from Helm v5.x to v7.x, or do we need to go through v6.x first? We’re currently on v5.11.2.
After restoring the MongoDB dump, should the already-accepted devices reconnect automatically? We’ve done a similar restore in the past (same Mender version) and it worked, but this time we’d be moving to a newer version — so just want to make sure.
Are there any DB schema migrations we need to run between Mender 3.7.7 and the version shipped with Helm v7.x? Or does the chart handle that automatically?
Does this plan sound reasonable? Anything we should watch out for?
Hi @rod.hpadilla ; actually it’s better to switch to the external MongoDB instance within the same Mender Server version, then upgrading later. Also, please consider the MongoDB versions involved for the dump and the restore, to avoid incompatibility issues.
Regarding the questions:
Can we jump directly from Helm v5.x to v7.x, or do we need to go through v6.x first?
Better to move in line and not skipping major versions; be mindful that Mender Server 4 is not compatible with the v5 chart, so you have to upgrade both the Helm Chart and Mender Server
After restoring the MongoDB dump, should the already-accepted devices reconnect automatically?
Yes, especially if you just replace the Bitnami database with the external MongoDB one, and you don’t yet upgrade Mender. Of course, it’s better to take backups and precautions as always. Make experiments on non production environments as well
Are there any DB schema migrations we need to run between Mender 3.7.7 and the version shipped with Helm v7.x? Or does the chart handle that automatically?
The charts handles them, as long as you keep the defaultdbmigration.enable: true