diff options
| author | Amrith Kumar <amrith@tesora.com> | 2016-12-16 09:49:49 -0500 |
|---|---|---|
| committer | amrith <amrith.kumar@gmail.com> | 2017-01-15 23:04:48 +0000 |
| commit | 6f5f107e3f319bc417c3de4a1a17616fea9060ba (patch) | |
| tree | 975388ad2b0a4c4d9e3c3266011b79975e59a3f8 | |
| parent | 523efe23140db10e376e747acb597ab6560a12b6 (diff) | |
| download | trove-6f5f107e3f319bc417c3de4a1a17616fea9060ba.tar.gz | |
backoff in poll_until
The current implementation of poll_until will attempt the retriever
every second and given that the call itself may take some time, this
effectively ends up being a non-existant sleep. On the under-powered
CI systems, this has an unwanted bad effect with polls running
continuously.
Changing poll_until to use the BackOffLoopingCall instead.
Change-Id: I9487d753e65c4de7753d9db19626497217c71f63
Depends-On: Id5be526a0418db5d00cbf5cbceb4993274989e72
| -rw-r--r-- | trove/common/utils.py | 6 |
1 files changed, 4 insertions, 2 deletions
diff --git a/trove/common/utils.py b/trove/common/utils.py index 52a3e7f8..d846b97b 100644 --- a/trove/common/utils.py +++ b/trove/common/utils.py @@ -199,8 +199,10 @@ def build_polling_task(retriever, condition=lambda value: value, if time_out is not None and time.time() - start_time > time_out: raise exception.PollTimeOut - return loopingcall.FixedIntervalLoopingCall( - f=poll_and_check).start(sleep_time, initial_delay=False) + return loopingcall.BackOffLoopingCall( + f=poll_and_check).start(initial_delay=False, + starting_interval=sleep_time, + max_interval=30, timeout=time_out) def poll_until(retriever, condition=lambda value: value, |