[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[leafnode-list] Re: Two (or more) systems synchronization

Martin schrieb:
> Hello,
> Am 07.10.2007 schrieb Angel Martin Alganza:
>>> First of all, each server gives article numbers. So when your leafnode
>>> finds a new article, it increases the current article number for the
>>> group, and gives it to the article. Your first task is: Keep the
>>> article numbers synchronized.   
>> Does that mean that for a given group article numbers never get used
>> again?  I mean, is there one and only one article number 1 even when
>> the first post gets erased?  
> Not quite right, as far as I know. Articles are counted and counted,
> and 1 is given, when
> 1) the spool has to be rebuilt (I do not know about something like this
> in leafnode, I don't think, situations like these exist, as far as you
> don't remove your whole spool)

A lost/damaged active file at the same time as an empty group might cause
the article counter to be reset in leafnode. This should rarely happen on
solid hardware with a solid OS.

> 2) the counter has an overflow. Since I know about article numbers with
> 5 digits, I assume, this case only occurs after 99999 or more articles.

The counter won't overflow until it hits 2^31, and I hope I fixed
everything so we actually have headroom until 2^32, or 2^64 for 64-bit code
on 64-bit and mixed 32/64 bit platforms.

> I'd try to rsync the whole /var/spool/news (or whereever your spool
> resides). leafnode stores some more information therein, and I believe,
> this is needed. Since I don't know about more information stored
> somehow in the file system (except /etc/, so take good care about
> similar configuration!), this should be enough.

Any variable data is in $spooldir (/var/spool/news), any static data in /etc.

> I hope, that all information needed by leafnode is stored
> within /var/spool/news. And I hope, there are no hard links. (That
> could double your needed space, if I see it correctly. More damage
> should not be done.)

Leafnode makes extensive use of hard links, and relies on them being
available, so you'll definitely want to recreate hard links rathe than
splitting them up; so it's rsync -aH or similar.

> I tried INN. It's nice if you have loads of time to spend. And if you
> want to play around at home. But otherwise... stick with leafnode. ;-)

Thanks for the flowers ;-)

> While you're at scripting: Just
> touch /var/spool/news/interesting.groups/* (or something like this, I
> haven't looked for the name, I get lazy sometimes ;-). Be aware, that
> you have to manually remove unread newsgroups then.
> (rm /var/spool/news/interesting.groups/gmane.leafnode.something)

It's even simpler: running fetchnews with the -n option :-)

> I think, most of the limits can be worked around. NNTP is not thought
> for configuration of news servers. But why shouldn't I just post an
> article to "my.leafnode.config" (this group is restricted and can only
> be used by me!) and change the whole configuration? Why shouldn't some
> lazy program just put its logfile in my.leafnode.logfiles? Yes, because
> NNTP is not intended to be used this way. But if that's the only
> reason ... why not?

I'd like to see the security concept surrounding these bits :-)

leafnode-list mailing list