This is a common question that we’ve encountered in the past during our PoC process.

Question:

How does the Knowledge Grid handle corruption?

Answer :

KG files are created during data load. They are managed safely in case of loading corruptions / other problems, i.e. automatic rollback should maintain the old KG files as well as data itself.

On the other hand, there is no special mechanism to control integrity and correctness of existing KG files. If damaged, they may be manually deleted and ICE still is able to answer queries, although maybe slower. There is also no mechnism to recreate / repair KG without reloading of data.

With the likes of Murphy and his associated law and as with any computer system, properly backing up the system and data is critically important. Although we could look at a adding a tool to detect/repair corrupted KN’s; we can’t necessarily detect or correct corrupted data packs. Under proper operation of the system, this sort of corruption should obviously not happen. But then there’s old Murphy just to prove me wrong.

So we would recommend regular back ups of the data files and if possible after every load. If you are loading intra-day, you might want to make sure the load files are safe and backed up so if the ICE back-up is missing some intra—day data, it can be reloaded.

Just a short note that some elements of KG are created during query execution not load.