On Fri, Jun 22, 2018, 15:37 Loren M. Lang <[email protected]> wrote:
> [...] but in the case of copying/rsyncing multiple backups from one drive > to another, those hardlinks are invaluable for space saving, but will > require much more RAM just due to the multiplied file count. Syncing 30 > backups means potentially 30 times the number of files to track. > Yeah, the hardlinks issue is one I encountered primarily when moving backups to a new drive. My solution there was to build a shell script to reperform the snapshots one by one from the old backups to the new one. This limited the memory required. Another thing I have done occasionally is to search all the content for duplicates and then hardlink the files where they weren't already hardlinked and enough metadata matched. This has savings particularly when the backup drive backs up different systems with lots of common system files. The process involves creating a map of pathname to inode number, computing the sha256sum of all the inodes, looking for common sha's with different inode numbers, and the relinking pathnames where the inodes differ, but metadata is sufficiently similar for your purpose. This process can take many hours and a series of manual stops. Also, spaces in pathnames can be a significant bitch and a half. However the space savings can be significant. > _______________________________________________ PLUG mailing list [email protected] http://lists.pdxlinux.org/mailman/listinfo/plug
