Is it safe to round-robin the nntp server?

Is it safe to round-robin the nntp server?

[Konstantin Ryabitsev]

Hello:

With lore.kernel.org now 3 different nodes with failover, I'm curious if it's
safe to point the NNTP server at that as well. I'm assuming that article
numbers are going to be the same across all three systems, but I wanted to
double-check that it's the case. I.e., if node1 goes down and DNS shifts to
point at node2, will that mess up nntp access, or is that okay? What about
imap?

-K

Re: Is it safe to round-robin the nntp server?

[Eric Wong]

Konstantin Ryabitsev <konstantin@linuxfoundation.org> wrote:
> Hello:
> 
> With lore.kernel.org now 3 different nodes with failover, I'm curious if it's
> safe to point the NNTP server at that as well. I'm assuming that article
> numbers are going to be the same across all three systems, but I wanted to
> double-check that it's the case. I.e., if node1 goes down and DNS shifts to
> point at node2, will that mess up nntp access, or is that okay? What about
> imap?

Likely, it depends on how picky NNTP clients are.  The one thing
that could differ is the "created_at" field in msgmap.sqlite3.

created_at probably affects IMAP more (UIDVALIDITY), though; and
I know some IMAP synchronization tools get picky about that.

Re: Is it safe to round-robin the nntp server?

[Konstantin Ryabitsev]

On Wed, Sep 01, 2021 at 08:54:13PM +0000, Eric Wong wrote:
> > With lore.kernel.org now 3 different nodes with failover, I'm curious if it's
> > safe to point the NNTP server at that as well. I'm assuming that article
> > numbers are going to be the same across all three systems, but I wanted to
> > double-check that it's the case. I.e., if node1 goes down and DNS shifts to
> > point at node2, will that mess up nntp access, or is that okay? What about
> > imap?
> 
> Likely, it depends on how picky NNTP clients are.  The one thing
> that could differ is the "created_at" field in msgmap.sqlite3.
> 
> created_at probably affects IMAP more (UIDVALIDITY), though; and
> I know some IMAP synchronization tools get picky about that.

Is that value taken from git repos, or generated during indexing? All 3 nodes
replicating from the same set of sources, but they do indeed run the index
hooks at slightly different times.

-K

Re: Is it safe to round-robin the nntp server?

[Eric Wong]

Konstantin Ryabitsev <konstantin@linuxfoundation.org> wrote:
> On Wed, Sep 01, 2021 at 08:54:13PM +0000, Eric Wong wrote:
> > Likely, it depends on how picky NNTP clients are.  The one thing
> > that could differ is the "created_at" field in msgmap.sqlite3.
>
> Is that value taken from git repos, or generated during indexing? All 3 nodes
> replicating from the same set of sources, but they do indeed run the index
> hooks at slightly different times.

Whenever public-inbox-init runs for v2 (first public-inbox-index for v1).

Re: Is it safe to round-robin the nntp server?

[Eric Wong]

Eric Wong <e@80x24.org> wrote:
> Konstantin Ryabitsev <konstantin@linuxfoundation.org> wrote:
> > On Wed, Sep 01, 2021 at 08:54:13PM +0000, Eric Wong wrote:
> > > Likely, it depends on how picky NNTP clients are.  The one thing
> > > that could differ is the "created_at" field in msgmap.sqlite3.
> >
> > Is that value taken from git repos, or generated during indexing? All 3 nodes
> > replicating from the same set of sources, but they do indeed run the index
> > hooks at slightly different times.
> 
> Whenever public-inbox-init runs for v2 (first public-inbox-index for v1).

Fwiw, it should be possible to automate ->created_at
synchronization using Last-Modified from /_/text/config/raw.

"lei add-external --mirror" or public-inbox-clone +index will
both set ->created_at automatically to match Last-Modified of
the origin:

  https://public-inbox.org/meta/20211012114705.383-10-e@80x24.org/
  ([PATCH 9/9] www: _/text/config/raw Last-Modified: is mm->created_at)