Delete objects if 1 node fail / how to definitely delete the data?

Hello, can you tell me: when I stop one node delete some data from another node up the first node data from second node copyed to first node but must be erase why data dont erase? is there are no information about deleted data in node?

We believe you are seeing what we refer to as a “phantom record.” Here is how this can happen:

If you take down one node, it still has a record of the object. So when the second node deletes the record, the first one will still have it. When the first node is brought up, it will bring the object back up. Since the second node has deleted it, there is no way it can know that it was supposed to be deleted.

This is a common problem in replicated databases. We are working on ways to make sure that when a record is deleted that it stays deleted.

Well I am also confused about it’s uncertain behavior for the management…Any suggestions…

This should not be a common issue, since the defragmenter will normally clear up the space. The best way people have found to definitely delete the data is to first “touch” the record and set a very short ttl. This means that even if the record should show up again, it will be beyond its ttl and thus be deleted again. It is not the best solution, but necessary when you must guarantee that a record be deleted.