[
https://issues.apache.org/jira/browse/CALCITE-1214?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15262743#comment-15262743
]
Josh Elser commented on CALCITE-1214:
-------------------------------------
bq. How do you know that renewal threads get closed? Should Avatica be using an
execution service for this kind of stuff, so we can use a small thread pool for
everything?
Drat. My thought process was that closing the AvaticaConnection should also
stop this thread (since the provided credentials are limited to the HttpClient
which is contained by a RemoteService which is contained by the Connection
instance. I forgot to connect this up.
bq. instances of AvaticaConnection, RemoteMeta, KerberosUtil (aka
KerberosConnection) and Service are in one-to-one (possibly optional
one-to-one) correspondence. If that is the case, they should all be able to
find out about each other, even though currently the "arrows" may point in the
wrong direction.
I have sort of started a precedent of using the AvaticaConnection instance as
the "discovery" point for this. While I don't necessarily like it, I always
don't want to commit to a re-work of this all before an avatica-1.8.0. How do
you feel about a bookmark to revisit this topic (proper
ordering/construction/sharing of objects used by an AvaticaConnection)?
> Support url-based kerberos login
> --------------------------------
>
> Key: CALCITE-1214
> URL: https://issues.apache.org/jira/browse/CALCITE-1214
> Project: Calcite
> Issue Type: Improvement
> Components: avatica
> Reporter: Josh Elser
> Assignee: Josh Elser
> Fix For: avatica-1.8.0
>
>
> I was talking with someone today about JDBC drivers in BI-tools/app-servers
> which reminded me about the general guidance to "make your driver work with a
> jar and a JDBC url".
> With the work I added recently to support client authentication with
> Kerberos, I wrote it under the expectation that clients are already logged in
> via a ticket cache (and have the appropriate system properties or environment
> variables set to point at the ticket cache). This doesn't jive with the
> original "jar and url" notion.
> It should be pretty easy to support a principal and keytab property to enable
> these types of environments.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
