----- Original Message -----
> On 10/05/16 00:55, Frank Ch. Eigler wrote:
> > ...
> > Until the $PCP_RC_DIR/pmcd script learns to delegate to systemctl, or
> > pcpqa tests switch to ".../pmcd reset" instead of stop, this neat bit
> > of systemd functionality is inappropriate for master.
Hmm, this seems readily solvable - one possible approach below.
>
> The world is not Linux and Linux is not systemd ... stop/start has
> worked forever in the pre-systemd world ... pcpqa is not going to change
> as this would be a massive disruption and break most things whenever
> systemd was _not_ in the picture.
>
> So the first option (change the "rc" scripts to be systemd aware when
> that is appropriate) is the correct approach I believe.
>
AIUI, this is a systemd feature that is not present on any other platform,
where it will automatically restart pmcd/pmproxy/pmmgr/pmwebd on "failure".
Optimistically, this assumes the failure will just go away on restart.
But anyway, this problematic interaction with QA could be resolved by the
addition of common.qa code that provides _{disable,enable}_auto_restart()
functions for use at the start/end of affected tests. The plain text unit
files are easily moved aside temporarily (as with pmcd.conf, pcp.conf, and
all manner of other config files) to provide the expected test environment
and in a way that is platform independent.
Also, thanks to Lukas and his band of buildbots for detecting this problem
early!
cheers.
--
Nathan
|