Updating a hosted map image layer with minimal downtime

[DORPAD]

Hello all,

I'm trying to find a better way to update a hosted map image layer. 

The current setup:

1)  Hosted map image layer published via ArcPro, with data reference from a feature class in a SDE.  The SDE is registered as a datastore.  The map image is only overwritten if there is a symbology change or something along those lines.

2)  The registered feature class is updated via a python script that essentially deletes out of date data from the feature class and inserts new data.

This works fine, but makes for awkward user experience during the time it takes to update the feature class, they layer doesn't draw at all, or at best only shows partial data.

I'd like to greatly reduce or entirely eliminate the downtime incurred by the update process.  It seems like this would be a pretty common issue with a well defined workflow, but I'm not asking google the right questions.  So I'll pose the question here. 

What's the best way to update a map image layer given the setup outlined above with minimal/no downtime?  Currently using Enterprise 11.3, ArcPro 3.2.

Thanks in advance

[JeanettaWyman_Timmons]

Hi

It sounds like your map image layer is not "hosted" but instead "referenced" since it is using an SDE database as its data source. 

Do you manually run the Python script and that is when the interruption in the service occurs? If so, you could schedule the script to run at a time when the service is likely not being used using Windows Task Scheduler.

[DORPAD]

Thanks for your reply!

The python script is run manually.  Scheduling it to run in off hours is problematic, our IT has been pretty aggressive in pushing out updates and rebooting machines.  

Since the last post, I have gone down the path of publishing the data as a hosted layer and overwriting that, and was pleasantly surprised to see that the overwrite only took about 6ish minutes, which is an acceptable amount of downtime for the service.

I was hesitant to try that as it is a pretty large feature class, ~1000000 polygons, 50ish attributes, but that's what I get for not trying the simple thing first.

Still open to hearing other approaches to solving this sort of problem though