I am aware there is currently no "comfortupdate" upgrade path for LS2.x to 3.x

We have a large number of running surveys to migrate from 2.64 to 3.7 we would prefer not to be "straddling" two systems particularly when many administrators will be running surveys in both this would cause a lot of confusion, never mind the whole URL issue.

So with this in mind we are wondering whether this pain is worth going through?

What are, if any, the compelling reasons to upgrade/migrate to LS3

For us:
* far better organisation of the administrative user interface (but not perfect yet)
* better ways of styling surveys through themes that work with inheritance, so you only have to add what is needed

Tammo

If you're not using custom plugins or themes, migrating from 2.7 to LS3 should be painless.

Unfortunately we are using customised themes. Some of them heavily customised. We are also using customised language files.

The main concern is the 140 or so currently running surveys.

Thats why I am looking to see if there is a "compelling" reason.

I could not find any guide to the migration even the "upgrading limesurvey" page only deals with the 1.+ to 2.+ version migration leaving us somewhat unsure of what to do.

Can we import the existing surveys inclusive of responses?
What other pain points are there?

While they are reasons, I would not call them compelling. Especially if there is a major disruption to services to implement the upgrade and test each survey is operating correctly.

Afaik: importing surveys with or without responsed will work, because the database has (hardly) changed.

Themes will have to be re-created.

If you have multiple surveys running, I would suggest setting up a new instance of LimeSurvey and start porting over whenever a survey is paused. After that invite respondent with the new URL and let the old version "fade out". Migrating with many surveys running, having to re-create themes on the fly could end up in a nightmare.

Just my 2 ct.

Tammo

We are specifically trying to avoid running multiple instances. The number of support calls from confused admins would drown us.

We have a large number of survey admins to go with the large number of surveys.

Also some of the surveys are ongoing so they are never paused.

I agree with Tammo. Although I am a proponent of version 3.x, I think updating with that many active surveys is high-risk. If you set up a new instance, you can develop and test your templates before either porting the surveys via .lsa export or upgrading the 2.x version.

Indeed, you should just hire a professional to create LS3 themes and manage your upgrade.

Oh. We would never try to do an in place upgrade. That's suicidal.

We would create a version 3 and then get our admins to test their individual surveys. If all good, shut down the server repeat the upgrade process and then restart.

So if the db is upgrade-able as implied in your reply, how is that done?

That would save us a lot of time importing users and permissions and surveys.

I think being able to upgrade the DB is critical to us because it really means all we have to worry about is the template issue. Which will be painful for a couple of the heavily modified ones but not insurmountable.

The language files are again a pain but not insurmountable.

I also want to start writing plugins. While its not critical, its certainly would be better if I did that on LS3.

I'll just ignore the professional jibe shall I?

Your situation is tricky. You will have to upgrade eventually, because support for 2.7x will stop (or already has?). So not doing the upgrade is not a solution, as with 140 current surveys I assume that you are using LS seriously.

I am always a fan of parallel installations for new major upgrades, because it saves me a lot of headaches. I can keep old surveys running until they are finished, I won't have sudden problems with existing surveys and I can thoroughly test new surveys.

But then I have a small number of people using LS actively (mainly me... ). With a big number of surveys and a big number of different survey admins such an upgrade can become very tricky. So I understand your point with not wanting to run a parallel installation, because of the big number of survey admins.

First of all, what I would do is to create a exact copy of my installation for doing the tests. Then I would do the upgrade. In my Version 2.73.1+171220 I get offered the upgrade to the latest Version 3.7.2 (180508), so Comfort Update would be an option.

Then I would do all the tests of the surveys, still working fine, etc. And I would create the Templates during this period as well.

Once everything is tested I would choose a time where you expect little survey traffic and do the comfort update. I know, with 140 ongoing surveys it is difficult to find the right time, but then there is never a time where nothing will happen. You will have to do it eventually.

Continuous surveys make it tricky. But depending on how you get people into those surveys, sometimes there are ways to guarantee that the flow is low (for some on the weekend, for some at night, etc.). Or if you send out invitations, then do it in between invitation waves. After 4-5 days when the invitations had been sent out, generally participation becomes very slow in my experience.

If you recruit on websites, just stop the recruitment process (probably javascript triggered) for a while and go for the upgrade then, etc.

Just a few ideas.