[BALUG-Admin] disruptions to some (mostly non-production) BALUG services

Michael Paoli Michael.Paoli@cal.berkeley.edu
Sat Mar 15 22:00:37 PDT 2014

Most of the immediate work on this has been done - and was in fact
"done" several hours ago, *however*:

before consistent service is restored.

Most notably at present, the services should generally "work", for
certain definitions of "work".  Most notably at the moment there are
two independent hosts providing the services.  And any changes that are
made on the "old" host may not be seen on the "new" host.  Also any
activities that may require consistent access to a single host may fail
or not work as expected.  But "other than that" the services should
generally "work" at this time.  Essentially I'm still waiting for
apparently at least 2 layers of DNS to get fully caught up, before all
should be fine and working essentially 100% again.

Technical details:
"old" host is/was on:
"new" host is on:
DNS changes have been made, as feasible and to the extent I have access.
However 2 things are delaying the DNS changes having full effect.
1) Hosting provider's DNS infrastructure - DNS changes have been made
    and submitted, however the submitted DNS changes don't take effect
    and propagate immediately - even with their "flush" function, which
    claims to update all their Internet facing DNS within 30 minutes - it
    lies, it doesn't - it's been many hours, and their nameservers aren't
    yet consistently showing the newer DNS data.
2) TTLs for many of the applicable DNS records are 14400 and I cannot
    change those TTLs (they're set by hosting provider and not customer
    changeable).  So, once 1) ceases to be an issues and all of hosting
    provider's DNS servers are serving up the newer DNS changes, at that
    point it time it will take up to 4 hours (the 14400 TTL) before those
    changes are then fully effective across The Internet.
At present I'm guestmating it should be "all better" by about
2014-03-16T04:00:00-0700 - but we shall see.  In any case, should be
within about 4 hours of whatever point in time 1) ceases to be an issue
for the DNS changes propagation.

Also, I'll probably move the host again in future to a higher
availability location - but expected that transition will be better
planned/coordinated and less disruptive.  Do continue to note that
present location of "new" host isn't particularly high availability -
so it *will* occasionally be unavailable (or quite slugish due to
bandwidth congestion).  However I expect it will generally be available
for the most part.

> From: "Michael Paoli" <Michael.Paoli@cal.berkeley.edu>
> Subject: disruptions to some (mostly non-production) BALUG services
> Date: Sat, 15 Mar 2014 11:01:10 -0700

> Due to needed physical host relocation (not sure of precise timing),
> I'm expecting we may have some disruptions and/or non-continuity of
> service for some BALUG services.  Guestimated the impact may be on the
> order of some hour(s) or (probably) less, or may involve some
> discontinuity/inconsistencies of service (e.g. updates/changes may not
> propagate to new host location).
> I'll provides updates when I'm reasonably able to, however note at the
> present time:
> In "worst case" I'll be able to provide update when the dust is settled
> and the move more-or-less completed, and in any case I'll provide update
> when we can expect things are relatively settled and not as likely to
> generally suffer more disruptions (mostly notably services should be
> restored to much more consistent again, though availability probably
> won't be as high as the current colo setup, as host won't be going to a
> colo).
> This WILL NOT impact:
> www.balug.org
> balug.org (itself)
> BALUG's lists
> It WILL impact:
> many non-production services and some ancillary production services,
> e.g.:
> archive.balug.org www.archive.balug.org
> wiki.balug.org www.wiki.balug.org
> secure.balug.org
> Technical details:
> If it's on it will be impacted (IP will change).

More information about the BALUG-Admin mailing list