Backups of our Canvas data

[mgudites]

I had to put in a support ticket with Instructure the other day because we had an instructor who blew away some content in a course, and we needed to get it back. The /undelete/ option did not cover what she needed in this case. Support's suggestion to me was to restore the content from the Beta environment, which I'm told gets cloned from Production every week. This brought up another larger conversation, which is, what happens if restoring the course from Beta to Prod isn't a viable solution (like, too much time has gone by) -- how do we initiate a restore from backup?

Their response was, and this is a direct quote from an email: Canvas doesn't backup your data. We provide Beta, the beta environment is overwritten with data from the production environment every Saturday. Any work or content you add to your beta environment will be overwritten every week.

Our data isn't being backed up? So if an instructor messes up their production course beyond repair on a Friday, then Beta gets overwritten on Saturday, then the instructor contacts us about it on Monday, we're just SOL? The easy answer is just "don't mess up the course" obviously, but let's face reality which is that faculty are going to be faculty and it's going to happen.

Does anyone else have experience with this? Is our data really not being backed up in a way where if we needed a course brought back from the dead, it could be done?

It feels ludicrous asking this question for what is a cloud-based solution, but if Canvas isn't going to do it, is there a way for us to take backups into our own hands and backup the data to our servers on-prem, or maybe to a Dropbox or something?

[Robbie_Grant]

 @mgudites ,

Support is correct in that data is not backed up.  Instructure can and will, for a fee, retrieve data if a school wanted to pay for the retrieval.  Remember, on the cloud, nothing is really ever gone, just much harder to get to.  With that said, if the instructor reset the course or deleted the course, you can go page through the Page Views as an admin, find the course ID and use the Admin Tools to restore the course.  I have actually had to do this on a number of occasions.

Hope this helps!

Robbie

[lekern]

It has been our experience that between the undelete function, the beta

instance which is overwritten each week, and the test instance that is

overwritten every 3 weeks (or now, I think it will be with the monthly

Production updates), we have good options for recovery across time.

Additionally, our process each quarter is that our SIS creates new course

shells for every course being offered, and the instructors copy the

materials from their previous offering of the same course. This means that

they have the archive of the course materials they have used in the past

available to them. Also, by turning on Canvas Commons, we have the

alternative space where an instructor can keep course materials or entire

courses that is separate from the Production instance. Of course, you can

also export course content into packages that are able to be re-imported.

Note that all of these options except for beta and test environments only

include the course content and not any of the enrollments, submissions, or

grades.

It takes some getting used to the idea that there aren't system-wide

backups that provide admins with access to discrete components beyond this,

but what we have found in practice is that this works well. We have yet to

irretrievably lose content that is needed.

On Wed, Aug 28, 2019 at 11:37 AM mgudites@hodges.edu <instructure@jiveon.com>

[jsavage2]

" Is our data really not being backed up in a way where if we needed a course brought back from the dead, it could be done?"

In a word, No.

This is one of the hardest things for users (and sometimes admins) to wrap their heads around in the move to the cloud.

In most cases, traditional backups were never intended to serve the use case you describe. They were intended to restore the system and provide business continuity in the case of a hardware or software failure. The economics of physical storage, though, dictated that regular full backups were expensive and periodic full backups with incremental snapshots were they way to go in most cases, yielding the familiar "weeklies," "nightlies," and "hourlies" that most of had love/hate relationships with for so many years.

A side effect of that was that we *could* save people from themselves on occasion, as long as whatever mistake they made was made in the right timeframe. If the data was written and overwritten all since the most recent backup, of course, there was no record. On a nightly backup schedule it was always perfectly possible to lose 23 hours and 59 minutes of work...which could be a lot of work for grad students finishing theses and dissertations, or faculty members preparing publications. But it was possible in some circumstances to get some things back.

But that was never the *purpose* of the backups. It was an off-label use and users were always advised (begged, cajoled, pleaded with, talked to sternly) to keep local backups of anything they really couldn't afford to be without.

In the cloud, though, the paradigm is different. Instead of keeping copies of old data to restore from in case of failure, distributed systems keep hundreds (or thousands, or tens of thousands) of copies of the current data available for immediate use in the event of failures, which are expected to happen all the time. The RPO for failure isn't "last night", it's "now". "Always on" High-Availability is a very different approach to data stewardship than traditional recovery from cold backups or failover.

Canvas (well, AWS) keeps multiple copies of your data for recovery and resiliency, which is the purpose of backups, but they don't keep a *historical* record of your data to do it. The cloud is much more robust (in many ways) than traditional server rooms, but the side effects of traditional practices like incremental backups are lost along with the practices themselves.

Between them, though, the undelete/restore function and the test and beta environments do provide much of the same benefit as traditional snapshots, depending on what your old backup schedule was. 

For end users, though, the same advice applies as always: If you really can't live without it, keep a local copy..