[volunteergrid2-l] Thoughts on catastrophic events and garbage collection/expiration

[Jody Harris]

As I was working on the Failure Overview, I started thinking about our
specification of "expire.override_lease_duration" to "31day":

In the case of a regional catastrophic event, it could very well be more
than 31 days before services are restored. At the cost of "efficiency" I
would like to submit that we set the expiration time much higher.

Here are some scenarios I had not thought of before:
  - forest/prairie fires: It is entirely conceivable that a data owner would
not be able to restore access for 90 or more days in these events.
  - catastrophic health event: It is possible that data owner could be
disabled for several months, even to the extent that they would not be able
to keep their clients online and updating their shares.

I submit that we should reconsider our recommended setting
for expire.override_lease_duration to a longer time.

90days?

180 days?

360 days?

As always, there are costs and advantages. The higher the number of days,
the more expensive it is for the grid to keep "dead" or discarded data.

Thoughts?

I'm going to add "Administrator incapacitation" to the Failures list.

jody
----
- Think carefully.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://tahoe-lafs.org/cgi-bin/mailman/private/volunteergrid2-l/attachments/20110202/893029f8/attachment.html>