You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on May 21, 2024. It is now read-only.
We are running aktualizr as part of a systemd service, and when a secondary ECU is tasked to shut down, the network.service service is first brought down, followed by the aktualizr service. When the aktualizr process receives the terminate signal, we call SecondaryTcpServer's stop() which starts a local connection to unblock the accept call in the run() function:
Due to the network service being down, the accept() call in the SecondaryTcpServer::run() function never gets unblocked, which causes systemd to forcefully terminate the process after a long wait time:
systemd[1]: Stopped Network Name Resolution.
systemd[1]: network.service: Deactivated successfully.
systemd[1]: Stopped Network Connectivity.
systemd[1]: aktualizr.service: State 'final-sigterm' timed out. Killing.
systemd[1]: aktualizr.service: Killing process 713 (aktualizr-agen) wit.
systemd[1]: aktualizr.service: Failed with result 'timeout'.
systemd[1]: aktualizr.service: Unit process 713 (aktualizr-agen) remain.
systemd[1]: Stopped aktualizr pipeline.
systemd[1]: Stopped target Basic System.
systemd[1]: Stopped target Path Units.
A fix was done on our end via changing the server socket to be non-blocking such that we poll accept() calls instead, but we would like to mainly know if this is an issue from our issue usage of aktualizr, if we are shutting down services in a different order than aktualizr's guidelines, or if there are working tests run on machines where the network is brought down first.
Due to the network service being down, the accept() call in the SecondaryTcpServer::run() function never gets unblocked, which causes systemd to forcefully terminate the process after a long wait time:
That sounds annoying indeed.
A fix was done on our end via changing the server socket to be non-blocking such that we poll accept() calls instead
That sounds great! Do you have a retry loop or some other mechanism to keep checking for a connection?
we would like to mainly know if this is an issue from our issue of aktualizr
No, I don't think you're doing anything particularly strange there. I think you're just doing something that hadn't been thought of before.
if we are shutting down services in a different order than aktualizr's guidelines
There are no guidelines, for better or worse!
if there are working tests run on machines where the network is brought down first.
Hi,
We are running aktualizr as part of a systemd service, and when a secondary ECU is tasked to shut down, the network.service service is first brought down, followed by the aktualizr service. When the aktualizr process receives the terminate signal, we call SecondaryTcpServer's
stop()
which starts a local connection to unblock the accept call in therun()
function:Due to the network service being down, the accept() call in the SecondaryTcpServer::run() function never gets unblocked, which causes systemd to forcefully terminate the process after a long wait time:
A fix was done on our end via changing the server socket to be non-blocking such that we poll accept() calls instead, but we would like to mainly know if this is an issue from our
issueusage of aktualizr, if we are shutting down services in a different order than aktualizr's guidelines, or if there are working tests run on machines where the network is brought down first.Here's a successful shut down for reference:
Thank you!
The text was updated successfully, but these errors were encountered: