Symptom: ArchivesSpace crashes with out of memory errors. Indexer held up on events.
I haven't looked into the underlying factors but we've found that (at least from) transferring records between repositories that duplicate event relationships can be created. If the transferred record is particularly "large" there can be thousands of duplicate event relationships which will crash the indexer as it attempts to retrieve and resolve them.
I created this plugin to delete the duplicate event relationships:
Cool, , we'll see if we can pinpoint where such duplication may have taken place. I wonder if adding a validation / database constraint so duplicate relationships are caught before commit would be useful. Will investigate and report back.
If you have the JSON for any sibling events (those linked to the same record), that might be useful. I'd also be interested to see the modified-times on these sibling events and even the relationships (via event_link_rlshp) – maybe then we could associate the duplication with a particular event/user workflow.
This investigation seems to have become dormant. Is this still a bug?
- yes, people are still reporting this as a problem. There was a discussion of it on the Users Group listserv between Megan Mummey of University of Kentucky and Chelsea Lodbell of Swarthmore College on August 2. Our tech support helped Chelsea with this problem via a support ticket using Mark Cooper's explanation and plugin. Chelsea passed that info. along to Megan. has knowledge of the ticket and might be able to give you more info., but it's essentially what Mark describes above.
Unfortunately, we can't replicate this. We would need some more information (logs, versions etc) about when it is occurring in order to take this any further.
Based on investigations that and have been doing related to events (particularly system-generated events), there is a larger issue here.