This is ususally due to an issue with the poller. Most likely either a resource limitation at the time which caused the poller to terminate without getting to the point where the RRD archive updates or an issue connecting to / querying the DB.
If you check /var/log you should see a cacti-poller error log which might provide some hints.
Keep in mind the specs used when deploying Cacti as it is in my opinion a notorious resource hog. My current Cacti install had to be moved from 768M to 1G recently as my graphs went from around 200 to 250, and I started seeing similar behaviour