Tips for transitioning SIS Import to Diffing Mode

[glparker]

Is anyone else making use of the recently introduced SIS Import Diffing Mode?    It sounds great, very much like I was used to using back in my Blackboard days of 'snapshot' mode.  

Some initial question

1)

I tend to send .zip files of CSV's to Canvas logically groups by function,

users.csv

courses.csv

terms.csv

accounts.csv

memberships.csv

It looks like diffing works by flagging a single file with a diff_id.   If that's the case, can I no longer bundle files?  

2)   My Memberships file contains only recently changed records (14 days in my case) for students and teachers.   It also combines them from multiple terms.   That is, all future terms, the current term, and two previous terms, are all included in a single membership.csv.   To take advantage of diffing mode, I expect I'd need to at least

- Break the file up into separate terms (membership-fall-2015.csv, membership-spring-2016.csv, etc...)

- Include ALL memberships for that term in the file.

3)  With Diffing mode, if a member of the diffing group is missing from the CSV, they are flagged as deleted in Canvas, correct?   Does this delete follow the regular rules of deletion?   That is, if I subsequently put them back in the course, will all their content remain as it was when they were deleted?

Please share your experiences migrating to diffing mode.  

[millerjm]

Hi  @glparker ​

This is an old posting, but I was wondering if you ever implemented the DIFF mode for your imports?  We are looking at implementing diff because our files are currently a large dump of all memberships and can sometimes be over 200K enrollments (includes adds, drops, and withdrawals). 

I'm a little concerned about users since it DELETES any user who isn't in the file.  I'd have a lot less worry if I were able to tell it not to delete anyone who isn't in the file but is in the diff group.  😕

I'd be interested in tips from anyone who has successfully implemented this...

Thanks!

Joni

[gerasin]

And more then 1 year later - anybody is willing to share experience?

I want to move to Diff mode for summer and would like to hear what other people done/doing/recommend

[millerjm]

Natalia,

We just transitioned to diffing mode for our enrollment files.  I'd be happy to chat with you sometime about the process I developed.

Joni

[khasson]

I know this is an old post, but we have experienced the good, bad and ugly with diffing mode!  I need to compare and understand options with the diffing and terms.   When the term changes and the diffing is on it deleted all the old courses and the students were still in a course for that deleted term!!!  (hence the ugly!)  Now to prevent this from ever happening again what are the options?  Please anyone help!!!!  

[lmsstaff]

I feel your pain.  I have been battling diffing issues for about 3 years now.  I would be happy to help, just reach out to me here and we can connect.

[CP_MICHAEL]

Hi,

This post doesn't look very active, but I wanted to share our experience migrating from batch mode to diffing mode and see if anyone has encountered this issue. We send a separate enrollment CSV files for each term and we're in the process of migrating from batch mode to diffing mode.

Our current plan is to migrate just the future term (non-active) to diffing mode and keep the current term running in batch mode to prevent any issues with current academic progress.

After implementing these changes in our beta and test Canvas environments, the enrollment behavior appears to be working correctly. However, for some reason our courses.csv is no longer working. When we try to upload the file we get the following error: “Unexpected error, ID: unknown”. The file works fine in production, but for some reason after implementing diffing mode the file fails. The first run of the courses.csv fails after the enrollments_future_term.csv file runs once. When uploading the courses.csv it fails in the GUI or via the API and we're not using batch mode or diffing mode for these.

The courses.csv includes courses for both terms, so I'm not sure if that's an issue yet. We have an active case open with Instructure, so hopefully we can get this resolved quickly.

Well, since thread is here to share experiences, I hope this is helpful to someone in the future, and if you've ever ran into this kind of bug let me know.

Thanks,
Michael