Problems deleting record with large number of children

[dc...@prosentient.com.au]

Hi all,

 

I ran into an issue today trying to delete a record with a large number of children in Fedora 4.7.5.

 

First, we used the “Delete” button under “Delete Resource” via the Web UI, but it generated the error “java.io.IOException: Tried to send an out-of-range integer as a 2-byte value: 35639”.

 

Second, we used cURL to try a DELETE by directly hitting the API endpoint, but we got a time out from our Load Balancer. Not surprising. However, what was surprising is that we tried to delete one of the child nodes after that (by hitting the server API endpoint from localhost rather than through the Load Balancer), and after 2 minutes it returned a 204 indicating success. However, the child node wasn’t deleted. I tried this a few times and the same thing happened each time. The API said the node was deleted, but it wasn’t. I restarted Tomcat and retried the DELETE using cURL and then it was successful.

 

I know it’s not much of a bug report, but I figured I’d share that and see if anyone has had similar experiences.

 

I don’t know if Fedora was trying to delete the 35000 child nodes in 1 Modeshape transaction or if some Java thread in Tomcat was just lost or what. It was a bit disturbing that Fedora was sending me 204 responses when it wasn’t actually deleting the content…

 

David Cook

Software Engineer

Prosentient Systems

72/330 Wattle St

Ultimo, NSW 2007

Australia

 

Office: 02 9212 0899

Online: 02 8005 0595

 

[Dan Pritts]

On 28 Aug 2020, at 3:07, dc...@prosentient.com.au wrote:
> I know it's not much of a bug report, but I figured I'd share that and see
> if anyone has had similar experiences.

I've had similar experiences, and solved them in similar ways.

Dan Pritts
ICPSR Computing & Network Services