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
Because runtest.sh output is cached and uploaded only every 15 seconds, when I have a task that initiates reboot, anything potentially useful that I might write to console (like "rebooting.." messafe) is likely to be kept in the cache and only shown to user (observer of the job in Beaker WUI) only after the reboot.
For example, if I have runtest.sh like this:
case $REBOOTCOUNT in
0)
echo >&2 "$(date -Isec) REBOOTCOUNT too low, rebooting to increase it..."
rstrnt-reboot
;;
*)
"$(date -Isec) REBOOTCOUNT is OK now, going on..."
;;
esac
When such task runs, depending on exact timing the "too low" message is likely to not be shown until after the reboot, when it will appear along with the rest of the log.
The scenario when this makes most sense is when the reboot takes non-trivial amount of time, and at the same time someone would like to keep an eye on the status of the test.
During the rebooting period, not seeing the "rebooting" message is counter-intuitive and it might seem like eg. the code got stuck earlier.
One instance of similar test is in-place major RHEL upgrade test, where the "reboot" is where we go to upgrade initramfs and perform the whole transaction --- this can easily take an hour.
The idea is that restraint could make another flush & upload outside of the 15-sec heartbeat:
either when a specific command (added by this RFE) is called (this option would run the risk of being "abused"),
or as part of rstrnt-reboot,
or when the daemon is shutting down.
The text was updated successfully, but these errors were encountered:
Because runtest.sh output is cached and uploaded only every 15 seconds, when I have a task that initiates reboot, anything potentially useful that I might write to console (like "rebooting.." messafe) is likely to be kept in the cache and only shown to user (observer of the job in Beaker WUI) only after the reboot.
For example, if I have runtest.sh like this:
When such task runs, depending on exact timing the "too low" message is likely to not be shown until after the reboot, when it will appear along with the rest of the log.
The scenario when this makes most sense is when the reboot takes non-trivial amount of time, and at the same time someone would like to keep an eye on the status of the test.
During the rebooting period, not seeing the "rebooting" message is counter-intuitive and it might seem like eg. the code got stuck earlier.
One instance of similar test is in-place major RHEL upgrade test, where the "reboot" is where we go to upgrade initramfs and perform the whole transaction --- this can easily take an hour.
The idea is that restraint could make another flush & upload outside of the 15-sec heartbeat:
The text was updated successfully, but these errors were encountered: