this patch fixes the issue, but I am not sure if it fixes it in the right place.

the hang is in shutdown in task_destroy, where it hangs on a "busy" backend. resetting the busy flag fixes it. But we have an delete of a backend followed by the add of the same again. Shouldn't the add, when it is completed reset the flag ?

this patch fixes the issue, but I am not sure if it fixes it in the right place.
the hang is in shutdown in task_destroy, where it hangs on a "busy" backend. resetting the busy flag fixes it. But we have an delete of a backend followed by the add of the same again. Shouldn't the add, when it is completed reset the flag ?

But it's the delete function that sets it busy, but it does not unset it after it's done its work. To me it makes sense the way it is.

And if I try and set it "not busy" when we create the backend it causes the server to crash during instance creation. Investigating the crash, but I feel this is the wrong direction to take. We'll see what I find...

maybe I was unclear, what I don't understand is:
- we add a backend
- we run an import task
- we delete the backend and add it again
- the shutdown fails to destroy a task: why does the reference to the deleted backend still exist, shouldn't the delete of the backend completely remove it ? it should be irrelevant what flags are set

Funny you ask that, I'm also looking into how the old flags are getting resurrected... Perhaps the task has a corrupted inst struct after it was destroyed. Investigating this....

Also setting the flags to "not busy" during backend creation does not fix the issue.

Yeah, so its the destroy function. This also fixes the problem:

+++ b/ldap/servers/slapd/back-ldbm/instance.c
@@ -392,7 +392,7 @@ ldbm_instance_destructor(void **arg)
     attrinfo_deletetree(inst);
     slapi_ch_free((void **)&inst->inst_dataversion);
     /* cache has already been destroyed */
-
+    inst->inst_flags = 0;
     slapi_ch_free((void **)&inst);
 }

This is still a hack fix though. Perhaps the delete instance function needs to destroy any tasks as well. Still investigating...

Needed to change how the import task destructor checked if the task was done or not.

Please review this patch:

3eb443b..bc6dbf1 master -> master

746abe7..2ef4e81 389-ds-base-1.3.7 -> 389-ds-base-1.3.7

80c8795..009800b 389-ds-base-1.3.6 -> 389-ds-base-1.3.6

389-ds-base is moving from Pagure to Github. This means that new issues and pull requests
will be accepted only in 389-ds-base's github repository.

This issue has been cloned to Github and is available here:
- https://github.com/389ds/389-ds-base/issues/2461

If you want to receive further updates on the issue, please navigate to the github issue
and click on subscribe button.

Thank you for understanding. We apologize for all inconvenience.