File changes propagating in older snapshots: what am I doing wrong?
Matthew Dillon
dillon at backplane.com
Sat Apr 8 10:59:54 PDT 2017
0x0000000000000000 is not a valid transaction id. Not sure how it wound up
in the snapshot list but that slave has not been synchronized from a master
yet and until it is there is no real snapshot.
-Matt
On Sat, Apr 8, 2017 at 9:55 AM, <tech_lists at mail.com> wrote:
> Hi,
>
> I have other information.. the behavior started to be 100% reproducible on
> my machine: every time I added a file I found it in the past snapshot. But
> it appeared after a delay, ranging from 1 to about 20 seconds!
>
> % hammer snapls /home/attic/
> Snapshots on /home/attic/ PFS#2
> Transaction ID Timestamp Note
> 0x0000000000000000 2017-04-06 09:57:34 CEST -
>
> % pwd
> /home/attic
>
> % touch a_new_file
> % ls .@@0x0000000000000000
> disk_stats.txt
>
> The file isn't in the snapshot yet. But if a i keep repeating the ls
> command after a few seconds I find it:
>
> % ls .@@0x0000000000000000
> a_new_file disk_stats.txt
>
>
> Then I decided to run hammer cleanup:
>
> # hammer cleanup
> [... Output snipped ...]
> cleanup /home/attic - handle PFS#2 using /var/hammer/home/attic
> snapshots - run
> prune - run
> rebalance - run..
> reblock - run....
> recopy - skip
> [... Output snipped ...]
>
> Now I have a new snapshot but the old snapshot is gone:
>
> % hammer snapls /home/attic/
> Snapshots on /home/attic/ PFS#2
> Transaction ID Timestamp Note
> 0x00000001186c2200 2017-04-08 18:25:50 CEST -
>
> This new snapshot doesn't exhibit the behavior described above: I can add
> and delete files in the main directory and the snapshot doesn't change.
>
> Sadly, now I have no access to the old snapshot to further diagnose the
> problem. I didn't think hammer cleanup would delete a snapshot from two
> days ago.
>
>
> Kind regards,
> Andrea
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.dragonflybsd.org/pipermail/users/attachments/20170408/ab94f87a/attachment-0003.htm>
More information about the Users
mailing list