Commit Graph

4 Commits

Author SHA1 Message Date
Vitaly Baranov
14a7ee8e26 Copy files to S3 during backup directly without using WriteBufferFromS3 to decrease memory consumption. 2023-01-17 09:35:41 +01:00
Vitaly Baranov
b9f2f17331 Add test and logging. 2022-11-01 12:23:20 +01:00
Vitaly Baranov
1365105bc4 Implement backup to S3 2022-10-19 00:04:41 +02:00
Azat Khuzhin
8830f0608d Support BACKUP to S3 with as-is path/data structure
Right now backup to S3 does not make a lot of sense, since:
- it has random names, and to decoding them
- requires metadata from local disk (/var/lib/disks/DISK/BACKUP_NAME)
- or send_metadata (but it is also tricky even with it)

So this patch adds simpler interface for S3, it is only suitable for
BACKUP/RESTORE, so don't try to use it for MergeTree engine.

It is done by adding separate disk - `s3_plain` for this, that:
- does not support any extended features, like renames/hardlinks/attrs/...
  (so basically everything that MergeTree requires)
- only write/read/unlink/list files

Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
2022-10-14 12:59:25 +02:00