As was mentioned on IRC, turns out it wasn't an MSC bug; the handler for
composite (i.e. multi-step) operations was blocking in an MSC callback
method[1], which is a no-no. So I'm fixing that.
[1]
http://fpaste.org/vcII/
On 2/22/11 12:47 PM, David M. Lloyd wrote:
On 02/22/2011 12:49 PM, Scott Marlow wrote:
> Here are some details about a failure to shutdown the standalone as7
> server. I'm using a private JPA branch, which was merged with
> master/trunk on Sunday.
>
> Deployment (NPE) error that preceded the failure to shutdown:
>
http://pastebin.com/xyjLQPbQ
>
> thread dump:
http://pastebin.com/bmREwigy
>
> as7 service dump:
http://pastebin.com/wRhx4deW
>
> pastebin links expire in 30 days.
>
> I fixed the NPE in my DUP but I'm wondering if we should protect against
> this at a higher level?
Looks like you found a bug in MSC. In that service dump, the INSTALL
phase of your deployment depends on all the other stuff in there, so
it's "holding" them all in the UP state. The problem is, once mode is
changed to REMOVE, that service should transition from START_FAILED to
REMOVING, but it didn't. I'll open an issue for it.
Thanks for thinking of doing the service dump.
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat