The chaos tests sometimes saw a message arrive twice after a pod kill or a drain (3 of about 22 chaos jobs). Each time the send had been retried after 421 4.3.0 ... Server shutting down. That 421 answered QUIT, after the end of DATA already had its 250: smtplib's with sends QUIT on exit and raises on anything but 221, so the lab took a delivered message for a failed one and sent it again. Reproduced against a fake server that answers QUIT with 421. The server did nothing wrong.
mailflow.py and soak-load.py now QUIT and close whatever the answer, and retry only what failed before the 250.
The soak running on the bench keeps the old sender (it isn't touched until it ends); a message twice in its report gets checked against its retries.
Tracker: inbuxa-charts#14
The chaos tests sometimes saw a message arrive twice after a pod kill or a drain (3 of about 22 chaos jobs). Each time the send had been retried after `421 4.3.0 ... Server shutting down`. That 421 answered QUIT, after the end of DATA already had its 250: smtplib's `with` sends QUIT on exit and raises on anything but 221, so the lab took a delivered message for a failed one and sent it again. Reproduced against a fake server that answers QUIT with 421. The server did nothing wrong.
`mailflow.py` and `soak-load.py` now QUIT and close whatever the answer, and retry only what failed before the 250.
The soak running on the bench keeps the old sender (it isn't touched until it ends); a message twice in its report gets checked against its retries.
Tracker: inbuxa-charts#14
The chaos tests saw a message arrive twice after a pod kill or a drain,
three times in about twenty runs. Each time the send had been retried after
"421 4.3.0 Server shutting down". That 421 was the answer to QUIT, after
the end of DATA had its 250: smtplib's with-statement sends QUIT on the way
out and raises on anything but 221, so the lab took a delivered message for
a failed one and sent it again. The server did nothing wrong.
mailflow.py and soak-load.py now QUIT and close whatever the answer, and
retry only what failed before the 250. The soak running now has the old
sender; a message twice in its report is checked against its retries.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
The chaos tests sometimes saw a message arrive twice after a pod kill or a drain (3 of about 22 chaos jobs). Each time the send had been retried after
421 4.3.0 ... Server shutting down. That 421 answered QUIT, after the end of DATA already had its 250: smtplib'swithsends QUIT on exit and raises on anything but 221, so the lab took a delivered message for a failed one and sent it again. Reproduced against a fake server that answers QUIT with 421. The server did nothing wrong.mailflow.pyandsoak-load.pynow QUIT and close whatever the answer, and retry only what failed before the 250.The soak running on the bench keeps the old sender (it isn't touched until it ends); a message twice in its report gets checked against its retries.
Tracker: inbuxa-charts#14