Error submitting or deleting items - URI too long when user is in a large number of groups

[ddean4040]

Describe the bug
When submitting or deleting an item, a user who belongs to a large number of groups but is not an admin will get an error message and the operation will fail.

The REST service logs:

2024-05-01 14:32:57,281 INFO  ... ... org.dspace.app.rest.utils.DSpaceAPIRequestLoggingFilter @ Before request [GET /server/api/discover/search] originated from /mydspace?configuration=workspace
2024-05-01 14:32:57,300 ERROR ... ... org.dspace.app.rest.exception.DSpaceApiExceptionControllerAdvice @ An exception has occurred (status:500)
org.apache.solr.client.solrj.impl.HttpSolrClient$RemoteSolrException: Error from server at http://solr:8983/solr/search: Expected mime type application/octet-stream but got text/html. <h1>Bad Message 414</h1><pre>reason: URI Too Long</pre>
        at org.apache.solr.client.solrj.impl.HttpSolrClient.executeMethod(HttpSolrClient.java:635) ~[solr-solrj-8.11.2.jar:8.11.2 ... - mdrob - 2022-06-13 11:27:56]
        at org.apache.solr.client.solrj.impl.HttpSolrClient.request(HttpSolrClient.java:266) ~[solr-solrj-8.11.2.jar:8.11.2 ... - mdrob - 2022-06-13 11:27:56]
        at org.apache.solr.client.solrj.impl.HttpSolrClient.request(HttpSolrClient.java:248) ~[solr-solrj-8.11.2.jar:8.11.2 ... - mdrob - 2022-06-13 11:27:56]
        at org.apache.solr.client.solrj.SolrRequest.process(SolrRequest.java:214) ~[solr-solrj-8.11.2.jar:8.11.2 ... - mdrob - 2022-06-13 11:27:56]
        at org.apache.solr.client.solrj.SolrClient.query(SolrClient.java:1003) ~[solr-solrj-8.11.2.jar:8.11.2 ... - mdrob - 2022-06-13 11:27:56]
        at org.apache.solr.client.solrj.SolrClient.query(SolrClient.java:1018) ~[solr-solrj-8.11.2.jar:8.11.2 ... - mdrob - 2022-06-13 11:27:56]
        at org.dspace.content.EntityTypeServiceImpl.getSubmitAuthorizedTypes(EntityTypeServiceImpl.java:151) ~[dspace-api-7.6.1.jar:7.6.1]

It looks like the query in EntityTypeServiceImpl.java is using GET instead of POST (it does not specify solrSearchCore.REQUEST_METHOD). We were able to workaround the issue for now by increasing the max header size in Solr.

This is on DSpace 7.6.1.

To Reproduce
Steps to reproduce the behavior:

1. Add a user to a large number of groups (our user is in 550 groups) but not the Administrator group
2. Try to submit or delete an item

Expected behavior
Operation completes successfully.

Related work
Looks similar to this old bug: #6732

[tdonohue]

This bug sounds similar to #9201 (possibly the same issue?) and both might be caused by #9164. We have a possible fix for #9164 in #9508 (and the PRs linked from that PR).

[ddean4040]

The URLs that threw 500s when our user ran into this issue were:

• /server/api/core/entitytypes/search/findAllByAuthorizedExternalSource
• /server/api/core/entitytypes/search/findAllByAuthorizedCollection

I haven't had any reports of slowness in the workflow tasks view, but I'll ask around.

[tdonohue]

@ddean4040 : Thanks for the additional details. Based on those details, I've determined this is a new issue (not a duplicate).

The core problem appears to be in EntityTypeServiceImpl.getSubmitAuthorizedTypes() (which is used by both of the endpoints you mentioned). That method appears to append every group you are a member of to a (potentially very large) Solr query. See the code at: https://github.com/DSpace/DSpace/blob/main/dspace-api/src/main/java/org/dspace/content/EntityTypeServiceImpl.java#L134-L139

So, this appears to be a very inefficient query, especially if you are a member of a large number of groups.

Pulling this over to our 8.0 board in search of a volunteer.

[tdonohue]

@ddean4040 : I just realized your analysis in the description is 100% correct. This problem should be fixed if we send this query to Solr via a POST instead of a GET. In other areas of DSpace, we are primarily using POST queries.

I'll create a small PR which should fix this.

[tdonohue]

Small PR which should fix the issue #9548. @ddean4040 : If you have a chance to verify/test this small fix, please do let us know your findings.

[alanorth]

I hadn't noticed these, but we do have a large repository with a number of users who belong to many groups. Sure enough, I see over a thousand in the past month or so (logs are zstd compressed):

$ zstdgrep -a 'URI Too Long' log/dspace.log-2024-04-* | wc -l
1423

I have tested the patch and left comments there.

---

BTW, there is a related "too many boolean clauses" error in Solr from other aspects of the UI, for example community and collection lists. I have one user who has 1,805 "OR" clauses!