By default the notifications are only allowed from the main process or any
child process that systemd finds. When the main process has not yet been
detected by systemd but MaxScale sends a notification, an error is logged
into the system journal. To prevent this, notifications from all member
processes should be allowed.
Any administrative operation, performed using maxadmin or maxctrl,
that takes longer than the watchdog timeout will cause MaxScale to
be killed and restarted by the watchdog.
By default ASAN only reports the errors and doesn't cause the program to
crash. By forcing a SIGABRT, the process will generate a core dump which
causes the test to fail.
Although the act of placing environment variables inside the systemd
service file might seem intrusive, they have no effect on non-ASAN
builds. This appears to be the most convenient and straightforward way of
changing ASAN behavior for testing.
Every time MaxScale was restarted, the test waited for a total of 25
seconds. As MaxScale should start before the command returns, the sleep is
not necessary.
Added StartLimitBurst=0 to the systemd service file to allow rapid
restarts of MaxScale.
If systemd restarts MaxScale when an abnormal exit is detected, it is
likely to happen again. This leads into a loop which causes multiple
maxscale processes on the same machine. One example of this behavior is
when systemd times MaxScale out when it is starting.
The maxscale service will be restarted only if the process is down because
a signal was caught or the process times out. If the process would be restarted
when it exits with a non-zero value, this would lead to infinite loops when
there is a configuration error.
The fact that the maxscale process exits only after all services have been
started means that systemd will kill the maxscale process if starting the
services takes too long. Since the user authentication data is loaded on
startup and can take up a long time, there needs to be a longer than
default timeout for systems that use systemd.
