[Borgbackup] Hostname change, now there is a problem?

[Thomas Waldmann]

> I added —files-cache=mtime,size to the borg create command, as I didn’t want to re-backup everything, since certainly all the inodes are different.

Assuming you did the previous backups with borg 1.1 and default options,
that should have been ctime,size (as the default is ctime,size,inode).

Changing to mtime likely caused some waste of time.

Changing back to ctime will waste some more time, but after that, it'll
be ok again.

> So, the initial run on the new server took over 11 hours, and looked like this:

Yeah, because likely the ctime in the cache was different from the mtime
read from the fs rather frequently, so it re-chunked a lot of files.

> Synchronizing chunks cache...
> Archives: 14, w/ cached Idx: 0, w/ outdated Idx: 0, w/o cached Idx: 14.
> Fetching and building archive index for kauai-2018-02-01 ...

^ Also that likely took a bit.

> The first thing I noticed that it didn’t list the previous retained archives with the old hostname,

Hm? Not sure what you mean.

Archive names (historically stored into the repo) do not change by
moving the repo.

> and then, the second borg run the next day was still running after 24 hours and I finally killed it.

Each time you change --files-cache value that may trigger a full
rechunking. So if you removed your mtime,size, it went back to the
default ctime,size,inode.

The inode part should be no problem as the first run updated the inodes
in the cache. But the mtime -> ctime change likely made it rechunk often.

> It didn’t seem to be transmitting any data,

Yes, because the chunks it made were already in the repo IF the file did
not really change.

> is my remote archive now hosed?

Well, the backup run you cancelled is of course incomplete, so maybe
there is a checkpoint archive and some after-commit stuff that will get
cleaned up automatically the next time borg uses the repo.

So, guess the best option now is to remove the --files-cache option to
go back to its defaults and just let it complete, even if it takes long.

You can remove the .checkpoint archive(s) AFTER you did a complete
backup. Prune should also do that automatically.

> Or is there a way I can salvage it or am I simply doing something wrong?

Your change to mtime was causing the waste of time, but everything else
should be fine.

borg 1.0 used mtime by default, but we switched to ctime in 1.1 for
better / more safe change detection.


-- 

GPG ID: 9F88FB52FAF7B393
GPG FP: 6D5B EF9A DD20 7580 5747 B70F 9F88 FB52 FAF7 B393