Cause: Database errors could occur if multiple node records, with and without an set target portal group tag, were created for the same iSCSI target portal. VDSM was doing this by creating static records after discovery.
Consequence: Depending on the filesystem dependent return order of the readdir syscall, sometimes a error would occur causing an update operation to fail.
Fix: Prevent multiple node records in different formats from existing simultaneously, detected at record creation time.
Result: Duplicate node entries should no longer exist in the iSCSI database, and updates to records should not result in a database error.