[ixpmanager] Automating the boring technical emails found on many IXP mailing lists ...

[Martin J. Levy]

Nick,

> ...  Do you know anyone who acts on them? ...

Sorry. I didn't address that issue. I assume the classic action is to
hit the delete button. I see zero use in them; however ... they exist
and I'm happy if they still existed; but ... maybe under control of
IXP (ie: subscribe/unsubscribe/digest/etc).

They do serve ONE major service (which is why I don't delete them; but
archive them). They provide a solid insight into a live member at a
specific peer. That can be very valuable when peering@ doesn't work.
Alas, maybe my recommendations would remove that useful feature.

> At INEX we monitor v4 sessions ...

I note that in your response you show a reliance on "bgp4-mibv2"; but
I'm very sure that the use of a non SNMP interaction between modules
could also be accepted. (Even an expect script generating a txt file
would be acceptable if not called to often). I would seriously think
you should separate function from communications method.

Martin


On Fri, Oct 11, 2013 at 10:11 AM, Nick Hilliard <nick at inex.ie> wrote:
> Hi Martin,
>
> We've been talking internally about some of these things for quite a while.
>
> - "we're announcing prefix $x": I figure these emails are pointless because
> no-one acts on them.  The only reason that people send them out is because
> other people send them out.  Do you know anyone who acts on them?  I kinda
> maintained some of them for a couple of months in the late 1990s, but
> that's about it.
>
> - maintenance: this is an area which we've been looking at for a while and
> where we have some definite ideas about how to make it work.
>
> - maxprefixes: currently we monitor this manually.  we plan to allow
> members to modify their own settings.  Automatic monitoring is a pain
> because we don't have bgp4-mibv2 reliably supported on the route servers.
>
> At INEX we monitor v4 sessions on LAN1 only on our route collector; we
> can't monitor v6 sessions and we can't monitor v4/lan2 because that's
> running bgp on a vrf-lite which needs bgp4-mibv2.  But this monitoring is
> slightly outside the scope of ixp manager.
>
> Nick
>
> On 11/10/2013 16:32, Martin J. Levy wrote:
>> I don't know if I should class this a feature request or just a subject
>> for discussion. (Maybe we can talk about this in Helsinki).
>>
>> I get (as an end-user of many IXPs) lots of emails from members with the
>> style I've included at the end. We can discuss many aspects of the value
>> of these messages; but they exist. Add to that the classic "We will be
>> taking down our router for maintenance at 23:00 ..." messages and we
>> have useful data; but sometimes its just classed as noise.
>>
>> I'd like to discuss how a central piece of s/w like IXP Manager could
>> handle these in a more automated manner.
>>
>> For example:
>>
>> 1) Use route server filtering to trigger these email's. Does one have to
>> assume IRR updates are always valid?
>> 2) Use a route announce page (via member login) to list new routes aka
>> the email below.
>> 3) Use a maintenance page (via member login) to "schedule" a peering
>> connection downtime. Use the route server BGP sessions (and/or ping
>> and/or SNMP) to measure the up/down times.
>> 4) Use route server BGP session review script to handle max-prefix
>> settings for member sessions. Allow the member to specific their expect
>> max prefixes.
>>
>> Each of these would generate an email; however ... two optimization can
>> be thought of. A) A digest of events. B) Different levels of information
>> to different per member email addresses.
>>
>> I'm thinking this could well be useful if there's a way to consolidate
>> IXPs or at least provide some consistently. For example some IXPs
>> mailing lists are full of emails like the one below; but others are not
>> even if the same ISP is connected to various IXPs.