Notify antivirus, on success, calls the process_virus_scan_passed taks.
Previously, this task would:
* update status to created (or delivered for test keys)
* copy the file from the scan bucket to either the live or test bucket
based on the results
* delete the old file in the scan bucket
We want it to:
* download file from scan bucket
* sanitise PDF using new template-preview functionality
* if sanitise failed, set to new status "validation-failed" and save
the pdf somewhere.
* send new pdf to live/test bucket
* update status to created (or delivered for test keys)
* delete the original file in the scan bucket
This PR does some of that:
* download file from scan bucket
* sanitise PDF using new template-preview functionality
* if sanitise failed, just log.
* send OLD pdf to live/test bucket
* update status to created (or delivered for test keys)
* delete the original file in the scan bucket
So if sanitising fails, we won't fall over and not deliver the letter,
we'll just log a message for now. If sanitise throws an unexpected
error (as opposed to a 400), we'll retry up to fifteen times (the same
as when creating a new letter). I've added the code for using the
sanitised pdf, but it's commented out for now
If there are no files to delete we won't get an excpetion.
Wrap the delete file in a try/except to avoid stopping the entire task.
Fix the missing slash for the file name.
rather than looping through, just use `next`. We only expect to return
one file, so this won't actually affect the code flow. If, for whatever
reason, the file isn't on s3, whereas previously it would error when
trying to return an uninstantiated variable, now it will raise a
StopIteration
The admin app still needs to use the name column.
Add the text field to the post data schemas.
If the text is not in the post data, then populate it with the data in the name field.
This should make the migration to text easier, and will work until we are able to update the admin app.
Celery prints its own stuff to stderr in a non-json format.
These get picked up by syslog drain and are shipped to
logstash together with the normal json-formatted logs resulting in
duplicate information and reducing the usability of the logs.
