mksnap_ffs woes

[Andrea Venturoli]

Kris Kennaway wrote:

> It's possible you're seeing deadlock bugs in the snapshot code; you'd
> need to compile your system with DDB support and obtain traceback
> information as described in the developers' handbook chapter on kernel
> debugging.

Ok. As soon as I get the chance, I'll do this (I cannot stop a whole 
network right away).





> The snapshot code was intended to support background fsck and that
> alone.  It's also optionally used by the dump code, but it was not
> written as a general-purpose live filesystem snapshot service.

Ok.
I just think that this, as well as the disclaimer about it being alpha 
code and the access locks, should be mentioned in the Handbook. Reading 
that chapter or mksnap_ffs's manual, I didn't get it and I suspect 
people might get the idea that it is stable code, which provides some 
functionality that it doesn't.




>>Are there other alternatives for live backups?
> 
> Plenty, if you don't demand an instantaneous image of the backup
> image.

That depends on what you mean for instantaneous...

I'll explain my needs briefly: I export (via Samba) a whole bunch of 
databases that 9 clients read and write. I do not have more details (I 
didn't write that app myself).
The needs are:
a) backup data so that in case of severe failure we can get it rapidly 
back (we are going to loose the last transactions, obviously);
b) replicate everything to another machine (via rsync) so that a remote 
user can see a snapshot of the system status.

We probably can live with some incoherent records in a) (after all, 
something will almost for sure need manual fix anyway); as for b) I fear 
it might be more prone to problems.

So, if for instantaneous you mean at a specific time, I don't care at 
all wether the image is made an hour earlier or later. All I'd like is 
coherence.

BTW, we have almost no room for changes on the client side :(





> The pros and cons of the snapshot code have been discussed on the
> mailing lists, and there is a (slightly outdated) information file
> here: /usr/src/sys/ufs/ffs/README.snapshot

I'm reading all this stuff; I might get back with some more questions 
later :)


  bye & Thanks
	av.