Engine hangs on unload

[lhaeger]

-------------- NSMDev wrote:

> It should unload - but on occasion there is a job that needs to finish so we can exit cleanly, which can sometimes take a while. However, we have seen cases where a thread just won't close properly and won't allow the engine to shut down. We are working on finding those issues so we can always unload fully, but as you might suspect, they are difficult to reproduce consistently so they are not easy to track down. We understand this is frustrating because to get the console prompt back, a server restart is necessary.
>
> The engine for 2.5 has been multi-thread enabled, thus allowing it to work through pending events much more quickly and efficiently. The screen counts you see are indicators of what types and the number of events that are in process. There are four types of worker threads that perform different functions, from left to right these are user storage, collaborative storage, auxiliary storage, and action objects. So where you have 1 | 0 | 0 | 0, that means there is one user storage job still not complete.
>
> I hope this helps a little

Thanks a lot for the GUI explanation, this is really interesting to know.

Looks like NSM got itself into a dead end in my case, one of the two pending events (probably the one being processed) ran into an error so it could never complete. But I could not abort it either, since the "unload nsmengin" command had already disabled the management interface.
Maybe these situations could be worked around by adjusting the order in which the various components/threads of NSM are unloaded. The mgmt API should probably be the last part to get stopped. Would that make sense?

Would it be helpful to send you the logs showing what happenend (debuggin was not enabled at that time, sorry)?

Cheers, lothar

--