i386 tests exhaust threads
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Gnocchi |
Fix Released
|
Medium
|
Julien Danjou | ||
2.1 |
Fix Released
|
Medium
|
Julien Danjou |
Bug Description
I was trying to backport gnocchi to the trusty-mitaka ubuntu cloud archive (which builds the package on i386 only). Tests are executed as part of the package build. After several tests have run successfully (~1300), the remaining 222 tests fail with the the same "thread.error: can't start new thread" failure. For example:
=======
FAIL: gnocchi.
gnocchi.
-------
_StringException: Empty attachments:
stderr
stdout
Traceback (most recent call last):
File "gnocchi/
return func(*args, **kwargs)
File "gnocchi/
super(
File "gnocchi/
return func(*args, **kwargs)
File "gnocchi/
super(RestTest, self).setUp()
File "gnocchi/
return func(*args, **kwargs)
File "gnocchi/
self.storage = storage.
File "gnocchi/
return get_driver_
File "gnocchi/
super(
File "gnocchi/
self.
File "/usr/lib/
_start_
thread.error: can't start new thread
Carbonara appears to be running out of threads. trusty python packages build on i386 only whereas xenial are on amd64, so that may be why this only surfaces on i386 for us.
description: | updated |
Changed in gnocchi: | |
status: | Fix Committed → Fix Released |
milestone: | none → 2.2.0 |
Makes sense, the heartbeater thread is never stopped. We need to implement a stop() can call it.