Upgrade Instructions for Minor and Major Versions

[Алексей Андреевич Шило]

Upgrade Instructions

This question concerns upgrades in a broad sense.
Goal: to compile a more or less complete and universal set of actions required when performing a new upgrade.

Problems:
1. The instructions for upgrading to the next minor version (e.g., 6.6 -> 6.7)
do not contain precise instructions for creating data backups.
2. Upgrading to major versions (e.g., 5.x -> 6.x) does not contain general instructions as such.

I'm basing this on what's stated in the base documentation. Quote:
https://guides.dataverse.org/en/latest/installation/upgrading.html

When upgrading within Dataverse Software 4.x, you will need to follow the upgrade instructions for each intermediate 4.x.
The same applies for 5.x -> 6.x.
Upgrades always involve deploying the latest war file but may also include updating the schema used by Solr or other manual steps. Please consult the release notes associated with each release at https://github.com/IQSS/dataverse/releases for more information.

For example:
https://github.com/IQSS/dataverse/releases/tag/v6.7

Upgrade Instructions
Upgrading requires a maintenance window and downtime. Please plan accordingly, create backups of your database, etc.
If you are running an earlier version, the only supported way to upgrade is to progress through the upgrades to all the releases in between before attempting the upgrade to this version.

Thus, we see that:
1. Upgrades require a mandatory sequence of actions (14 steps in the case of v6.7), performed manually.
2. The upgrade process requires a period of maintenance and downtime.
3. The upgrade process has its own specifics for each intermediate version.

Okay, let's assume this is inevitable. But I'd still like to clarify the phrase: "create backups of your database, etc.", especially "etc."
I'm asking those who have experience with multiple and successful upgrades from intermediate versions to write down the sequence of commands that must be executed to guarantee NOT losing data.

Now regarding the second issue.
There was a post here on February 29, 2024, about upgrading to the major version.
https://groups.google.com/g/dataverse-community/c/BUO37-reWIs/m/18SK0DZlAQAJ

Here's the quote:
However, to upgrade from 4.2.2 to 6.1, there are 60 intermediate versions. Not yet mentioning the changes in the versions of PostgreSQL, Solr, Java, and Glassfish4 to Payara6 between the three major dataverse versions. We also applied customization to the code. We will need to customize and test each version upgrade. If we follow the official upgrade guide, the upgrade project from 4.2.2 to 6.1 will take us a few months to complete.

The author then proposed a 9-step strategy, which they applied to their project.
There was one reply to this message, calling the strategy an innovative approach. And that's it.
I'm likely missing something, but the official documentation hasn't changed since then and hasn't been updated with a new strategy.

I'd like a professional answer. Can we use this strategy to skip at least a few intermediate versions, say, 6.6 - 6.11? Or is that still at our own risk?

[Philip Durbin]

Hi! Thank you for your thoughtful message.

If by professional, you mean "by someone from the core Dataverse team", I'm happy to give this a shot.

https://guides.dataverse.org/en/6.11/installation/upgrading.html (which is identical to the latest version you linked to, as of this writing) is still correct. The officially supported way to upgrade is to march through each upgrade one by one. This is what we do at Harvard Dataverse. Obviously, doing so is a huge pain for installations that are way behind, such as the 60 intermediate versions you mentioned. We'd love to make this easier. It's a hard problem to solve. We desperately want all installations to stay up to date, especially for security.

In the absence of the core team having the bandwidth to make upgrading easier, the community has stepped with approaches such as the one you found from Franky. Let me try to enumerate all community-led upgrade approaches I'm aware of:

- Franky: https://groups.google.com/g/dataverse-community/c/BUO37-reWIs/m/18SK0DZlAQAJ

- Don Richards: https://github.com/gdcc/dataverse-recipes/pull/25

- Oliver: https://github.com/gdcc/dataverse-recipes/pull/33

There may be others! I hope others jump in with any I forgot. Yes, approaches like the above are at your own risk.

We know upgrading is hard. Again, we don't feel the pain as acutely at Harvard Dataverse because we upgrade four times a year, with every release. The pain is spread out.

I hope this helps! I'd love to hear what others think about this topic as well! How can we make upgrading easier?

Thanks,

Phil

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-commu...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dataverse-community/22ef515f-7c62-43d6-a8cc-03b46d07b09fn%40googlegroups.com.

--

Philip Durbin
Software Developer for http://dataverse.org
http://www.iq.harvard.edu/people/philip-durbin

[Olivier R]

Hi,

Just want to say that Franky's strategy worked for me to migrate from 5.10 to 6.7.1. However, this was our "sandbox" Dataverse with just a few small datasets. Currently. I try to do the same with our true Dataverse repository, which might be another story. Still too soon to say.

I'd like to hear other feedback as well.

Best regards,

Olivier

IPGP Datacenter