Last week I wrote about a duplication error we’d found that was causing jobs to fail with an (incorrect) “Duplicate URL” error. To summarise the problem… if a submission timed out after the actual post had been made but before the success check had been completed then the job would be rescheduled to be retried a few hours later but woud then fail on the second attempt with a duplication error because the original post had in fact succeeded.
We implemented a fix for that and if a duplication error is found, we now just skip ahead to the part where we pull out the URL of the new bookmark that was created. However, we hit a snag there… we use a large pool of accounts for submissions and we use them at random for each job so when a job was retried there was no guarantee that the same account would be used the second time and in this case we wouldn’t be able to pull out the Bookmark URL as it would be on somebody else’s account.
So the solution for that was to start tracking which account is used with each submission job and then in the case of a retry (which happens on server errors) it will use the same account for any subsequent retries of that particular job to ensure we can correctly grab the Bookmark URL that was created.
We have implemented this fix today but please note that it is not retroactive. Meaning, we cannot go back into past jobs and figure out what account was used so we can only start recording this information from this moment onward so for any jobs that were submitted into the system prior to this fix, it’s still possible that you will get some jobs that are not correctly reporting the Bookmark URL that was created.
All new jobs going forward though should now be working fully as intended!