Re: [Tools-discuss] Server Transition Briefing

[Russ Housley <[e-mail]>]

Here is a summary of the discussion on the call that took place on March 11th.

Russ


= = = = = = = =


IETF Tools Team Discussion with the Community on 11 March 2020


The IETF Tools Team held a one-hour WebEx session on the server
transition that took place on 22 February 2020.  The goal was to share
information about the changes in the server architecture over the last
decade and the most recent upgrade.  Then, with this context in mind,
hold a discussion with the community.

The slides about the server history were sent out in advance, and they
can be found in the mail archive at:

https://mailarchive.ietf.org/arch/msg/tools-discuss/X5Wy2WlA5oUV7SoNSm2Fson5tDU/

Some questions got the discussion started after the briefing.  The first
two questions were discussed together.

Q1. How can the approach to monitoring/alerting be more transparent?

Q2. How to ensure monitoring/alerting covers the end user experience?

It was suggested that the IETF LLC pay for monitoring at status.io,
pingdom.com, or newrelic.com to provide everyone with information about
whether the IETF services are up or down, the percent of availability
over time, and so on.

It was suggested that the IETF LLC pay for Splunk reviews of the server
logs to help find performance issues.  For example, right before the
Internet-Draft cut-off date, review of the logs might highlight ways
to improve the user experience.

There was a desire to automate spot checking of email messages flowing
through mailman into the mail archive.

Q3. How to improve the infrastructure for testing?

It was suggested that all of the traffic against the system be collected
for a day, then use that to determine whether changes are improving the
user experience or otherwise.

Q4. What are the next steps toward separation of services to reduce
fragility and increase flexibility?

It was suggested that the IETF leverage things that are available in the
market wherever possible.  That is, avoid building custom software where
possible.

It was suggested that public cloud services should be used instead of
private cloud services.

It was suggested that microservices be explored.  An OpenID Connect or
OAuth 2.0 identity provider could be used to enable microservices.  Some
things that might be migrated to microservices are the Internet-Draft
repository and the Datatracker database.

Wrap Up: Was this discussion useful?  If so, how often should we hold a
similar discussion?

People felt the discussion was useful, and it was suggested that they
be held quarterly.