Server keeps reloading empty OD data

[wsanchez@…]

We have an optimization in the server which we keep a timed cache of OpenDirectory data for a while, but in order to avoid wiping our data store out when OD hiccups, we don't reset the cache timer if we get back an empty set for a record type.

In cases where there really is no data for that record type, this means the timed cache isn't used, even though is should.

[cdaboo@…]

PyOpenDirectory now raises exception when there is a an error, so in theory we could rely on that to determinethe error state. However, we need to check that when using the /Search node, if a particular LDAP node is unreachable, that an error is generated rather than it returning results for all the other working nodes. We need to make sure that a failure in any node explicitly included in the /Search results in a failure for the whole.

Another option might be to go with an explicit LDAP node rather than /Search. The one benefit of that is that it avoids problems with figuring out whether users from different LDAP servers need to be in different "realms" on the calendar server.

[wsanchez@…]

r1030

[wsanchez@…]

Milestone Preview 1 deleted