Does hot backup to s3 support other S3 compatible storages except Amazon s3 and MinIO?

Hi Dear,

I am trying the hot backup to s3 feature, but my S3 is neither Amazon s3 nor MinIO, Does it work? Thank you very much.

1 Like

Provided your S3 follows the S3 standard, it should work. What are you using?

1 Like

Thank you so much for your reply.
I can backup to s3 now.
And sorry to bother you, I have some new questions:

  1. How to restore from s3? The guide just only show how to restore from local dir. Is there some command to restore? or I need copy the data from s3 to local data dir manually?
  2. How to restore to a shard? Is it the same with the method of restoring to a replica set and data will be balanced later?
  3. Can we backup and restore by db or collection and backup with specified timestamp?
  4. Can we divide large data when backup or restore?
  5. Do we backup the oplog generated during the backup processing?

Your help is highly appreciated!!! Thank you so much.

1 Like

Hello @lily1 ,

  1. You just do pbm restore and it is going to use the backup configuration you have set before. In case of S3 it is going to download the files from S3. You don’t need to download files manually.
  2. It is the same method, yes.
  3. Unfortunately it is not possible to backup/restore single db or collection for now. We have it in our roadmap though. As for timestamp - yes. It can be done with point-in-time recovery being enabled.
  4. Not sure what you mean here. Could you please elaborate?
  5. Yes, we do.

Hi spronin,

Thank you so much for your so quick reply.
Regarding the first question, I am using percona  server for mongodb, and backup the data file by issue the command " db.runCommand({createBackup:1,xxxxxx})",  what is the corresponding command to the restore?  Seems pbm is logical backup, while the hot backup is physical backup, right? If I am correct, Does the hot backup support the PITR?
May I know the plan of the backup/restore single db or collection, I am looking forward to the feature. :-)
Regarding the fourth question above, I mean maybe even a single data file is large, I am wondering if there are some resource limit, Is there possible that the backup/restore process will failed if the whole file was copied. Is there some measures to avoid this, like divide a large data file into some smaller data files?

Thank you so much.

@lily1 , yes, PBM is logical and has support for PITR.
There is no PITR for physical backup yet, but we plan to implement it in PBM as well. ETA would be Q1’2022.

Restoring single collection is also planned, but this would be possible only with logical backup.

Regarding the fourth question above, I mean maybe even a single data file is large, I am wondering if there are some resource limit, Is there possible that the backup/restore process will failed if the whole file was copied. Is there some measures to avoid this, like divide a large data file into some smaller data files?

I assume it is for the case of hot backup? Let me check internally. @Andrew_Pogrebnoi any thoughts?

Hi spronin,
Very glad to get your reply.
So it can not be supported to restore single collection for physical backup, right? How about single db? It can not be supported for physical backup, either?
Regarding the restoring single collection for logical backup, seems mongodump supports it, What’s the difference between them?
Yes, it is for the case of hot backup. Looking forward to your reply.
Your help is highly appreciated!!!

Hello @lily1 ,

looks like I brought confusion instead of some clarity :slight_smile:

Solution Physical backup Logical backup PITR Single collection recovery
PBM No, but in the roadmap Yes Logical: yes
Physical: planned
No, but in the roadmap for logical backup
hot backup Yes, only one replica set No, not planned No, not planned No, not planned
$backupCursor Yes, sharded cluster as well No, not planned No, will be in PBM No, not planned

Hope this helps :slight_smile:

Hi spronin,
Thank you so much. It is very clear. :slight_smile: