My database size is over 400GB and it takes 16 hours to complete the full backup. I see PBM disable the PITR before the full backup start and resume only after the full backup finishes.
This is an issue as if the system crashes in between full backup, we can not restore the MongoDB to the point in time before the crash as we don’t have oplog chunks available.
We had other methods to backup oplogs separately and were able to apply oplogs manually but I think PBM should not be used for big databases.
On the one hand, I see what you mean. Yes, we might rethink the current approach on backup + PITR interaction. On the other, 400Gb is not the biggest database and for me, 16h is way too much time for such an amount of data. The primary constraint for the backup is the storage writes bandwidth. As reads from mongo during the backup are relatively cheap. Although the node load may have an impact, I’d take a closer look at your network and backup storage as the main suspects.
Thanks for raising the issue.
Right now, no PITR during backup.
You can track development status of two possible solutions for your case: