EBS Snapshot Backup(AWS EBS) Problems in PSA's Sharded Cluster

Ops and Admin
, ,
1

Hi:

We have encountered some problems in designing the backup scheme for the community edition of mongodb sharded cluster we built, and would like to ask one and all for advice on the following scenario, which is as follows:

The size of the sharded cluster (PSA’s sharded cluster) is relatively large, so we plan to use snapshots (snapshot service provided by cloud service providers such as AWS EBS) instead of mongodump/mongorestore, for mongodb cluster backup. However, we understand that ebs is difficult to ensure the consistency of the time of all the snapshots of the shard, and the production environment can not stop writing during the backup, that is, I understand that you can not use db.fsyncLock().

My questions are:

  1. What other options are available for the above scenario to ensure successful implementation of ebs snapshot backup (snapshot start time consistency)?

  2. If I use db.fsyncLock() in the SECONDARY of each shard, but the problem with this is that if the PRIMARY goes down during the snapshot, it would be catastrophic (PSA), not sure if I’m understanding this correctly?

If any more info is needed please say.

All the best folks!