Migrating from TrueCharts

[Protopia]

@Captain_Morgan

The decision to move to Docker was entirely based on community requests. There was a desire for Docker Compose simplicity and for greater reliability/support of Apps. Loosely coupled 3rd party Apps were getting many complaints for complexity, reliability and support challenges. it was the least favorite part of TrueNAS and the most common request to fix. It was negatively impacting TrueNAS reputation and business.

I am not questioning the decision to move to Docker. I am questioning whether the implementation route and the communications were as sympathetic as they could have been.

@Captain_Morgan

We both knew that mixing Docker Compose and Helm Charts was not feasible due to incompatibilities.

@Captain_Morgan

We used a 2 step parallel running process:

1. We supported jailmaker and nspawn containers (sandboxes) in Cobia. This allowed docker compose apps to run in parallel with Helm charts in Cobia and then Dragonfish. After these were successful, we made the decision to switch in Electric Eel to Docker Compose and alerted the Community well in advance. Dragonfish could run with Helm Charts for years as has been proven by 13.0.
2. In Electric Eel we also decided to migrate the Helm Charts Apps we controlled to Docker Compose Apps automatically (after much sweat, that is working). We also committed to running Kubernetes within the nspawn containers. This is still possible. 3rd party catalogs could have used this to support existing Helm Charts.

1. Since you admit that it is possible to run Kubernetes within an nspawn container, and it is also feasible to run Docker within an nspawn container, that means that is WAS feasible to run both the old Helm-based UI-driven apps infrastructure alongside the new Docker-based UI-driven apps infrastructure.

2. A two step migration (Helm UI predefined apps-> sandbox CLI manually defined apps → Docker UI predefined apps) is much much much more complex and sudden-death than parallel running of Helm and Docker both with a UI. It is only for those users with suitable skills and the time to spare - and completely infeasible for non-technical users who installed TrueCharts Apps in good faith on the basis that iX had encouraged 3rd parties to provide alternative catalogues.

@Captain_Morgan

3rd party catalogs had a choice, but we don’t control them.

1. iX provided infrastructure that allowed them to publish a Helm-based catalogue and then you pulled the plug. What exactly was their choice?

2. iX encouraged 3rd parties to invest time and effort into creating a Helm-based catalogue because it was in iX’s interests to have a much wider base of apps which would encourage TrueNAS take-up, and after TrueCharts had invested that time and effort and given iX the benefit of a broader apps catalogue, you pulled the plug without providing either TrueCharts or their users with any realistic migration path (by which I mean an iX provided easy non-technical UI-based non-sudden-death-cutover migration path). iX could have supported parallel running for a major release or two (yes - there would have been a bit more engineering effort), but you decided not to.

Had you provided parallel running for a major release or two, then:

• Migrations could have been done one-by-one whilst the server was in production, rather than as a sudden-death cut-over during the version upgrade to EE.
• TrueCharts (if they wanted to - and if not an alternative Community effort) would have had time to use the same migration functionality you created for the iX Chart apps to allow for migration of TrueCharts apps to Docker.

@Captain_Morgan

There are already more systems using Docker Compose Apps in Electric Eel than systems using 3rd party Helm Charts in Dragonfish. The adoption rate of Electric Eel is 3X higher than previously, even with the teething issues of migrating Helm Charts to Docker Compose. We’ll keep on polishing. By January, the Community will be well placed to pass judgement on our efforts.

This is NOT a reasonable comparison. Docker apps in EE include iX apps, 3rd party Helm Charts is just TrueCharts.

The users who have already migrated are likely to be at the more technical end of the user spectrum, and they are well placed to be able to achieve this smoothly.

But (as I previously warned would happen - but iX new better) we are already seeing examples of users who are upgrading and losing their apps, either because they were not aware that non-iX apps would simply disappear or because the iX-apps migration didn’t work for them e.g.:

• Need help,after upgrade to 24.10, all app missing, k8s_to_docker.migrate job failed
• Missing apps after upgrade 24.10

So please let’s not crow about how great the EE take up has been when there are users whose production services have been disrupted by an incomplete and apparently flaky app migration process that resulted from a decision by iX to do a sudden-death cut-over.

@Captain_Morgan

We’ll keep on polishing. By January, the Community will be well placed to pass judgement on our efforts.

Tell that to the non-technical (and admittedly perhaps somewhat naive) users who upgraded and lost their apps and now have to flounder around to try to get things working again. I suspect that they have already passed judgement on iX’s efforts.