One of the most noticeable features we included in this iteration is the addition of SHA-1 checksum files with all SCP “push” restores. After we push over the data files or the .tar.gz, we push over a file of these SHA-1 checksums. This means that you can know two vital pieces of information, that your restore is complete (since the checksum file arrives last), and that your restore is not corrupted in transit.
This iteration we also released new critical functionality for customers who are backing up MongoDB deployments. Currently, the MongoDB Backups done by MMS are accomplished via streaming oplog and periodic snapshots and clustershots of the data. When taking these ‘moment in time’ captures, however, we were always mindful of the MongoDB Balancer. The balancer is responsible for moving data between shards. If the balancer was on, a snapshot could not be taken until the balancer had completed its tasks and the data was stationary.
After today’s release, customers running MongoDB in a sharded cluster will be able to take clustershots even if the balancer cannot be stopped. Since the balancer is responsible for ‘moving the data around’ these clustershots can be inconsistent (have missing or redundant data), but customers with active balancers will appreciate being able to have this added level of availability and peace of mind knowing that at any moment they have a copy of their data (albeit an incomplete one). To take a clustershot when the balancer is on, customers will need to update their backup agent to the latest version, 220.127.116.11. Also in this iteration for MongoDB backups is restore file size estimations in the MMS web UI.
Monitoring MongoDB is priority one for our users, and this iteration we added a couple of key features. For the activity feed, we added an indicator to show the type and priority of a host. This is so that in case a “host is down” alert is generated, users can see whether that host is a primary or secondary, delayed, an arbiter, etc. Users can also now view deleted alerts in their activity feed as a method of auditing user actions.