Closed ansell closed 6 years ago
The ANDS DOI service may be back again (possibly due to an upstream issue with DataCite), and the status URL now returns the following:
{
"response": {
"responsecode": "MT090",
"verbosemessage": "(took 1ms)",
"message": "The rocket is ready to blast off -- all systems are go!",
"type": "success",
"code": 200,
"doi": "",
"url": "",
"app_id": ""
}
}
The DataCite status page which shows intermittent outages is:
Download service accepts an optional mintDoi boolean parameter which defaults to false for backwards compatibility and can be used to bypass doi minting.
Downloads plugin however seems to be always setting that to true and currently there is no way to make it configurable.
This isn't a user interface issue. We still want all downloads to have DOIs. The core issue is that the download must not fail if the DOI service isn't available.
Working on first part.
[9:51 AM] Peter Ansell: The core is that we need a failsafe where we are still able to deliver downloads successfully to users when DOI registration fails. [9:52 AM] Peter Ansell: One part of that will definitely need to be adding code to biocache-service to catch the failure of the DOI registration, log it, and continue, possibly with a different email template that lets them know the DOI is not available for that download. [9:54 AM] Peter Ansell: Another part may be adding some sort of retry mechanism to doi-service internally and block on the response for a bit longer before returning, and verify that biocache-service is able to sit and wait for the response. Or you could make it a callback to avoid relying on the HTTPS/TCP connection staying open, and biocache-service configurably knows how long to wait before making the download succeed without a DOI
Closing this based on Javier's testing of the code that switches back to using /data/biocache-download if the preferred DOI method fails, so when DataCite is broken ALA users can still get access to downloads.
We can discuss more fancy options like asynchronous retries to get a DOI in the future.
The ANDS DOI service is broken right now, which means that all downloads through biocache-clustered are broken. All services have been reverted back to biocache-service-1.x (biocache-lb) until the situation is fixed.
We need to have a failsafe in biocache-service for when this occurs. In particular, downloads must never fail if the DOI service is unavailable and the download would have otherwise succeeded.
I verified that the ANDS DOI status using the following URL, based on looking through the doi.ala.org.au sourcecode.
https://services.ands.org.au/doi/1.1/status.json
The log messages on doi.ala.org.au were:
The log messages on prod-bdown-b5 were: