We use the levelDB repository for a month out of two, which are now being produced on NFS (with the standard failover rollback installed). Over the past few weeks, we had a damaged storage, with no errors in the logs ... only piling up queues and very low bandwidth. The only thing we could do to solve this is throw the store away and start over.
So, we are back to the old and reliable KahaDB store.
source share