Re: dump won't dump an fss snapshot on netbsd-5

[Juergen Hannken-Illjes <hannken%eis.cs.tu-bs.de@localhost>]

On Tue, Jul 28, 2009 at 11:14:19AM -0400, Louis Guillaume wrote:
> 
> 
> Juergen Hannken-Illjes wrote:
> >On Mon, Jul 27, 2009 at 05:28:34PM -0400, Louis Guillaume wrote:
> >>Hi Juergen,
> >>
> >>Juergen Hannken-Illjes wrote:
> >>>On Sun, Jul 26, 2009 at 10:13:09PM -0400, Louis Guillaume wrote:
> >>>>I just updated the machine connected to my tape drive from netbsd-4 
> >>>>(stable) to netbsd-5 (stable). Now dump will not back up my fss 
> >>>>snapshots...
> >>>The /dev/fss? devices will not allow multiple opens.  To dump a snapshot
> >>>please use the -x / -X flags.  In your example it would be:
> >>>
> >>>   dump -L maat-home -b 32 -0eautf /dev/nrst0 -x /var/tmp/maat-home /home
> >>>
> >>
> >>This appears to work properly, however I have no indication that a 
> >>snapshot was made or even used aside from this:
> >>
> >>  [16:42:52 EDT] DUMP: Dumping a snapshot of /dev/rld0g (/home) to 
> >>/dev/nrst0
> >>
> >>Is it safe to assume that I don't have to create the snapshot with 
> >>fssconfig first? And that dump takes care of that for us? It's not 
> >>really clear in dump(8).
> >
> >Yes and yes.  You could run "fssconfig -lv" and "systat vmstat" during
> >dump to see a fss device working ...
> >
> >>Another thing - fss in general (both via dump and by itself) is 
> >>considered "experimental" but I haven't had much trouble with this until 
> >>now. Is this because there has been limited testing? Limited resources 
> >>to maintain/develop it? Perhaps it's not as stable on ports other than 
> >>i386.
> >
> >Limited testing. While I'm running snapshot dumps daily on all my
> >NetBSD machines I'm not aware of others using it on a regular basis.
> >
> >>I consider this a very important feature and would like to see it 
> >>considered "stable". It would also be good to get rid of some ironic 
> >>(albeit funny) documentation in the manual, such as:
> >>
> >>  -x snap-backup
> >>     Use a snapshot with snap-backup as backup for this dump.  See
> >>     fss(4) for more details.  Snapshot support is experimental.  Be
> >>     sure you have a backup before you use it.
> >>
> >>... from the dump(8) page.
> >
> >This is NOT meant ironic -- be sure you have a (non-snapshot) dump before
> >you try it.
> 
> 
> 
> Jeurgen - Thanks again. I do have one more curiosity...
> 
> Did something change in netbsd-5 that would break my old system of 
> manually creating the snapshots with fssconfig, then backing them up?

Looked at the source again and found myself wrong:
- fss devices allow multiple opens.
- with 5.0 block devices deny multiple opens.

You were dumping the block device whose behaviour changed in 5.0.

It should still be possible to dump the raw device with a command like

        dump ... /dev/rfss0

> While I am happy to do things the "right" way going forward, I have to 
> wonder how come this worked doing it the other way in the past.
> 
> I've dumped/restored fss snapshots for years without dump -x. Actually, 
> you helped me with this back in '05...
> 
> http://archive.netbsd.se/?ml=netbsd-users&a=2005-03&m=703385
> 
> ... Based on that I'm guessing that there was a change. Did we lose some 
> flexibility here? The way it is now, I don't think I can make a snapshot 
> now, and dump it later on when some other conditions are right. Not that 
> I really need that, but it seems like a possible scenario.
> 
> At any rate - I now have good confidence in my backup system once again. 
> dump works beautifully with -x. My backup script is simpler now too.
> 
> Please let me know if I can provide any testing or feedback that could 
> help to get fss out of the experimental stage...

-- 
Juergen Hannken-Illjes - hannken%eis.cs.tu-bs.de@localhost - TU Braunschweig 
(Germany)