All MP's children functions have the ability to detect if there is a problem and shut down's its own process. However, we need to use the future system which can detect if the MP child was aborted and also have a result passed back from the child MP process. You can then use that to update any lists that were feed into the Pool the first place.
Two problems are triggered without this fix.
- A list feed into a multi proc will need to be updated to remove failed child MP, especially if that list continues to used in more processing.
- It is possible for all MP children to fail. If all fail, we need to shut the whole app down.
PS. we already have some examples where we do use the 'futures' system. We just need to check if they need to update any lists for failed MP children.
Update: Mar 1 - Put this on hold.
All MP's children functions have the ability to detect if there is a problem and shut down's its own process. However, we need to use the future system which can detect if the MP child was aborted and also have a result passed back from the child MP process. You can then use that to update any lists that were feed into the Pool the first place.
Two problems are triggered without this fix.
PS. we already have some examples where we do use the 'futures' system. We just need to check if they need to update any lists for failed MP children.
Update: Mar 1 - Put this on hold.