On Mon, Sep 30, 2013 at 11:57:52AM +1300, Tim Foster wrote:
On 09/28/13 04:24 AM, Erik Trauschke wrote:
Because there's no dependency on the system-repository-cache, it
does mean that we can have this situation:
# svcs *system-repository*
STATE STIME FMRI
disabled 11:23:18 svc:/application/pkg/system-repository-cache:default
online 11:01:33 svc:/application/pkg/system-repository:default
where a rogue admin has disabled the cache service and nothing
decides the system is broken.
for the system-repository-cache service why don't we:
- deliver it as always enabled
- mark the dependency on the system-repository service as
grouping=optional_all and restart_on='refresh'
then when the system-repository-cache service start method runs we know
- the system-repository is online, in which case we start htcacheclean
- the system-repository is disabled, in which return
that way htcacheclean will get automatically enabled whenever the
system-repository service gets started.
[pkg-discuss] Re: Code review: system-repository service split