rmdmattingly commented on code in PR #5868:
URL: https://github.com/apache/hbase/pull/5868#discussion_r1607192892
##########
hbase-backup/src/main/java/org/apache/hadoop/hbase/backup/impl/BackupManager.java:
##########
@@ -295,51 +298,31 @@ public ArrayList<BackupImage> getAncestors(BackupInfo
backupInfo) throws IOExcep
.withRootDir(backup.getBackupRootDir()).withTableList(backup.getTableNames())
.withStartTime(backup.getStartTs()).withCompleteTime(backup.getCompleteTs()).build();
- // Only direct ancestors for a backup are required and not entire
history of backup for this
- // table resulting in verifying all of the previous backups which is
unnecessary and backup
- // paths need not be valid beyond the lifetime of a backup.
- //
- // RootDir is way of grouping a single backup including one full and
many incremental backups
+ // If the image has a different rootDir, it cannot be an ancestor.
if (!image.getRootDir().equals(backupInfo.getBackupRootDir())) {
continue;
}
- // add the full backup image as an ancestor until the last incremental
backup
- if (backup.getType().equals(BackupType.FULL)) {
- // check the backup image coverage, if previous image could be covered
by the newer ones,
- // then no need to add
- if (!BackupManifest.canCoverImage(ancestors, image)) {
- ancestors.add(image);
- }
+ // The ancestors consist of the most recent FULL backups that cover the
list of tables
+ // required in the new backup and all INCREMENTAL backups that came
after one of those FULL
+ // backups.
+ if (backup.getType().equals(BackupType.INCREMENTAL)) {
Review Comment:
I'm not so certain that we made the right call here. Experimenting with
incremental backups at my day job, and taking an incremental backup for one
table versus another (on the same cluster) results in totally different set of
file contents for each. I'm still not an expert in this code from beginning to
end, but we might have been too quick to assume that the API was lying
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
