a85e6e07cf
If a driver.spawn fails without triggering a re-schedule then the placement cleanup and the compute build failure counter update happens after every externally observable change (instance state change, instance fault recording, notification sending) in the finally block of _locked_do_build_and_run_instance() call. We have couple of functional tests that assert either the placement cleanup or the state of the build counter. These tests are unstable as the test races with the compute manager code. As there are no externally visible change to wait for this patch introduces some retry to these tests to stabilize them. Closes-Bug: #1870385 Change-Id: I68369e7fa7630a212f4b20fc09f4c40796934bb9 |
||
---|---|---|
.. | ||
functional | ||
unit | ||
__init__.py | ||
fixtures.py | ||
json_ref.py |