Allready twice we had a sudden increase of data in the epersongroup2eperson table. Acctually it was an exponential growth were the amount of data seems to double each time the table is accessed. Theoretically the table can hold 10.000 rows. It ended up with 65 mill.. And our repository was de facto down.
In 2017 we just cleaned the table with help of an backup. After that our dspace-repository had no further problems here. But as the problem arose again 2018 we decided to investigate a bit further.
The problem seems to be correlate to an increase in self-registering. (Once a year a whole lot of students upload their thesis by self-registering to our repository at the end of a semester.)
Unfortunatly we wer not able to pin it down to a code bug. But we decided as a quickfix to alter the epersongroup2eperson table to protect itself. The pair (eperson_group_id, eperson_id) should be declared unique so that the grpud to members realtion can protect itself from those errors.
ALTER TABLE epersongroup2eperson ADD CONSTRAINT one_to_one UNIQUE (eperson_group_id, eperson_id);
Now we are looking forward to 2019 .