-
Notifications
You must be signed in to change notification settings - Fork 400
Description
In order to implement cluster backup we need to improve box.backup:
- It should return the requested checkpoint and all the xlogs after the requested checkpoint, so that the backup is collected at the time "NOW".
- It should rotate the latest xlog, so that it isn't appended to during backup.
- It should return the timestamp of the last entry present in backed up xlogs. This timestamp will be used as a backup timestamp.
We should add a new box.backup.info() handle, which'll return the same info as box.backup.start() for an already started backup, or an error "Backup is not in progress".
We should make it possible for any instance of the replica set to recover from a backup made from any other instance of the replica set.
Optional, not requested by any customer yet: we should make it possible to recover to the middle of xlogs, up to the provided timestamp, with truncation of the remaining operations.
JIRA issues:
https://jira.vk.team/browse/TNTP-2825
https://jira.vk.team/browse/TNTP-3389
https://jira.vk.team/browse/TNTP-3391
https://jira.vk.team/browse/TNTP-3392
https://jira.vk.team/browse/TNTP-3406