On Wed, 17 Jun 2009, Ethan Quach wrote:

> Alok Aggarwal wrote:
>>>> - Also, since this smf service is separate from the the auto-install smf 
>>>> service,
>>>> how would the output/error from this service be reported?  Will it be a 
>>>> separate
>>>> log file from the auto-install service?
>>> 
>>> Its a separate log file.
>> 
>> One of the things the AI client redesign proposes
>> is to consolidate all the installation related
>> logging into the auto-install service log. A central
>> log location like this helps users in not having to
>> look at multiple places to determine what's going on.
>> 
>> It seems we're moving in the opposite direction by
>> creating another log file the user needs to keep
>> track of.
>
> As per our discussion on this at the meeting, I asked Liane
> about adding additional log file pointers to service definitions,
> and there currently isn't a proper way to do this.  In a service
> definition, there are tags to define doc links and manpage
> references that do get displayed with svcs -x, but these are
> metadata of the service, and not really fit for specifying the
> log file.  If we really want to be able to this, we could file an
> rfe.
>
>
> With the goal of having a single install log file in mind, we
> have a couple of options
>
> 1.  define the single install log file to be a separate file, and
> have each of the SMF service log files note where the install
> log is in their content.  This imposes an additional step for the
> user to -- look in smf log to get install log path, then go to
> actual install log -- but it keeps the functional separation of
> the services and provides for the proper granularity of restart
> and dependencies of the functionality we've defined.  Also
> when the rfe is fulfilled, we'll be able to note the separate log
> file in the svcs -x output.
>
> 2.  combine what we've functionally proposed as the manifest
> discovery service back into the auto-installer service.  We
> obviously lose benefits described above for #1.
>
>
> Please provides opinions/thoughts.  I would like closure on
> this by today if possible.

#1 is better than the alternative. Especially given that the 
deficiency in SMF doesn't allow us to satisfy the requirement 
of following the svcs -x output for diagnostic purposes unless 
we go with a single service.

So, I guess it boils down to the question of whether we want
modularity by sacrificing serviceability a small amount.

Alok

Reply via email to