- 3,734,328 Users
- 2,246,950 Discussions
- 7,857,228 Comments
- 17.8K All Categories
- 3.3K Industry Applications
- 3.2K Intelligent Advisor
- 57 Insurance
- 533.3K On-Premises Infrastructure
- 137.3K Analytics Software
- 38.5K Application Development Software
- 5.1K Cloud Platform
- 109K Database Software
- 17.5K Enterprise Manager
- 8.8K Hardware
- 70.7K Infrastructure Software
- 105K Integration
- 41.5K Security Software
Are backups of volumes on an OCI instance taken at or close to the same time?
When a backup is taken, what are the recovery-times of the volumes and can they every be skewed enough to prevent using that backup for a restore?
Normal backup policy (Bronze, Gold, etc) caution you to idle the disks during backup. For how long?
If I instruct OCI to backup an instance, will there be any variance in the timestamps for multiple volumes comprising the instance? If so, how much?
If an OCI instance has a file system built inside an LVM logical volume on a group that spans multiple OCI volumes, will the recovered disks always be consistent? If consistency relies upon blocking writes then for how long?
If these backups of disks composing a system are not taken close enough in time, is it possible for one of the member disks to be backed up with sufficient delay that the volume group could be damaged, requiring repair.
If one disk is updated but another disk needs to be updated to complete the operation and a backup is taken the resulting backup would have inconsistent, possible unusable disks.
An obvious disaster scenario is a backup during an active move of physical extents (PEs) between disks. Lost PEs are lost data, duplicate PEs result in corrupt data. Another problematic case is during a filesystem resize where many dirent and inodes are being rewritten. Many files could be lost or the superblock corrupted on the filesystem.
In a detailed example: your Oracle database is using the volume group datavg. The is logical volumes and the group spans /dev/xdb1 and /dev/xdc1.
Take a backup of the instance. If /dev/xdb1 backups up to a recovery-point (RPO) at 12:00:00am and /dev/xdc1 has an RPO later at 12:00:01am there could be inconsistency. Obviously if there is active I/O one would expect some problems,
If we need to restore /dev/xdb1 and /dev/xdc1, we don't have proof the backups were done as an "atomic transaction" or snapshot. (Yes, snapshots are not copies so they are not backups.)