Due to limitations of SQL Server and the method that Vault uses to handle transactions/operations, etc., deadlocks are a frequent occurrence when working in an environment with multiple job processors. If the Vault job queue would AUTOMATICALLY retry failed 'server busy' jobs, it would resolve the majority of these issues.
We use 6 different machines as job processors, as we often change state on entire packages of drawings at a time. It is very common for 50+ drawings to hit the job queue at once, leading to fairly frequent database deadlocks and lots of "Server Busy" errors in the job queue.
99% of the time, when a user resubmits the failed jobs from the queue, they complete just fine the second time through.
This is an annoying thing for users to have to do and could very easily be automated within the job queue.
If a user does not diligently check the queue and doesn't wait for certain jobs to be completed before moving files to another state, etc., this can cause all sorts of 'non-tip version' errors, files that are out of sync, etc.
Here's the idea:
In Global Settings, give us some additional settings:
Maximum failed job retries: Allow us to set this to 1-5 or something reasonable. The job queue will automatically add a failed job back to the queue, as the user that originally submitted it, exactly as if they selected it in the queue and hit 'Resubmit'. Take away the manual human interaction.
Retry delay interval: Range between 1 second to several minutes. The job queue will resubmit the failed job after waiting this long between retries.
Checkbox list for types of errors to resubmit for: Some types of errors will still fail if resubmitted automatically. Either allow the admin to select the type of errors that re-submissions will occur for or ONLY resubmit for "Server Busy" type failures.
This simple change would dramatically reduce the number of jobs a user has to manually resubmit. Since it appears impossible to completely weed out SQL Server deadlocks, implementing an automatic retry would make the majority of these deadlocks completely transparent to the users, ensuring that their files are published as expected and files stay in sync.
