On 4/4/25 08:26, John Doherty wrote:
Hi, yesterday I made a mistake and I'm not sure how to fix it without
resorting to --nuke (which would not actually be catastrophic).
I have one machine, let's call it A, that has been running tarsnap from cron
for a pretty long time. Yesterday, I made a replica of this machine, B,
without stopping to think about the implications for tarsnap. Since B is
mostly a replica of A, they both have the same keys (it's not strictly a
replica: they have different IP addresses and hostnames).
They both tried to run tarsnap at the same time. B succeeded and A failed with
"tarsnap: Sequence number mismatch creating checkpoint: Run --fsck", which is
understandable.
The current state of affairs is that B can run tarsnap --fsck but A cannot. If
I try, A reports:
tarsnap: Sequence number mismatch creating checkpoint: Run --fsck
tarsnap: Error fscking archives
tarsnap: Error exit delayed from previous errors.
The desired state of affairs is that A can run tarsnap again and B does not
need to. I've disabled the cron job on B, so it won't run again unexpectedly.
Not sure what to do. Maybe rename the tarsnap-cache directory on A and let it
create a new one?
Easiest option here is probably to delete the tarsnap-cache directory on A and
copy the tarsnap-cache directory from B over to A.
The other option is to delete the directory and run `tarsnap --fsck` to create
it again; but that involves downloading metadata from the tarsnap service so
there's not much point doing that when you already have a copy of everything
on the B system.
--
Colin Percival
FreeBSD Release Engineering Lead & EC2 platform maintainer
Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid
