we use storage level snapshot to recover from an outage for replica set. The outage is something like a power failure of all the nodes in the replica set. After we recover from the outage by mount the snapshot as it is, we found the count() method against all the nodes in the replica set gave a wrong result. Only after we use validate() method, then the count() can give the correct number. So our questions are:
1. if validate() method a must after recover from an outage of replica set?
2. is there anything we need to do besides validate() method?
3. shall we leverage fsynclock() just before the snapshot is taken?