Adding support for printing to work with a shared volume (cloud environment)

[tiago-s-vieira-alb]

Before this change, print-lib generates a file into disk, and cache the file pointer into a Java map (in memory).
The problem of this, is after generating a file, the next request to get the file MUST be to the same geoserver instance. Other geoservers, doesn't have any reference to this generated file. In a k8s environment this is a problem.

Before implementation:

• Not cloud environment and not shared volume, the loadbalancer must have sticky session --> it works
• Not cloud environment and shared volume, the loadbalancer must have sticky session --> it works
• Not cloud environment and shared volume, if loadbalancer do not have sticky session --> it will crash when get from a different geoserver then the one that generates the file
• Cloud environment, the loadbalancer cannot have sticky session --> it will crash when get from a different geoserver then the one that generates the file

To solve it:

• When in printing phase, instead of java map, we persist the information into a JSON file on disk (next to the PDF file)
• When requesting the file, the code now reads the JSON file from the disk and always provides the response
• The purge of old files (more than 10 minutes) was maintained

IN SUMMARY:
What was saved in an instance into an hashmap, now is persisted on disk.

After releasing version with this PR, create a ticket in geoserver like this "https://osgeo-org.atlassian.net/browse/GEOS-11480" to update mapfish-print version.

[tiago-s-vieira-alb]

Hi @jodygarnett , can you please review PR?
It will be great it could be integrated into a next minor geoserver version.
Thanks

[jodygarnett]

Okay! Yeah many other geoserver extensions end up with a shared database (rather than a shared json) to communicate state across instances.

I wish we had a good shared module for that which could use an appropriate storage / communication to avoid all the duplication we are seeing.

[jodygarnett]

Can I ask you to include a note on the docs about this as it is a pretty important changed :)

[tiago-s-vieira-alb]

Can I ask you to include a note on the docs about this as it is a pretty important changed :)

@ritika-t-thakur-alb updated documentation.

[jodygarnett]

So how does this work in a could environment - don’t they need to share a temp folder or something?

[ritika-t-thakur-alb]

Hello @jodygarnett ,
I elaborated my documentation a bit more.
Answering your question: depending on the cloud environment, engineers an pick an option to share a volume between multiple instances and pass the path in the existing MAPFISH_PDF_FOLDER environment variable.
I have already updated installation.rst file with the documentation for MAPFISH_PDF_FOLDER.

I think the steps for creating a shared volume is not related with this PR.

[tiago-s-vieira-alb]

Hi @jodygarnett, i think everything is done from our side.
Do you need more information?

[jodygarnett]

If you rebase to resolve conflict we should be able to merge?

[tiago-s-vieira-alb]

Hi @jodygarnett i think we don't have any conflict.
Can you check please?

[jodygarnett]

Okay I could not rebase but I can do a squash and merge

[tiago-s-vieira-alb]

@jodygarnett when you release print-2.3.2, please update also in geoserver:
https://osgeo-org.atlassian.net/browse/GEOS-11480

Thank you :)

[jodygarnett]

I think it is all done