[Freeipa-devel] [PATCH] 1074 limit service list

[Rob Crittenden]

Simo Sorce wrote:
> On Tue, 2012-12-04 at 15:14 -0500, Rob Crittenden wrote:
>> Simo Sorce wrote:
>>> On Tue, 2012-12-04 at 14:03 -0500, Rob Crittenden wrote:
>>>> Only touch the service list in the server installer and ipactl.
>>>
>>> Nack, comments inline.
>>>
> [..]
>>> This break the fallback we have in ipa_stop()
>>> We expect an exception or a non empty list there.
>>
>> Ok, I can move the handling so ipactl ignores the exception.
>
> Is the problem that we are printing an error to stdout/stderr ?
> Or do you actually want to change behavior somehow ?

We need to change the behavior.

If you run: ipactl stop

then: ipactl status

you get a backtrace because the service list doesn't exist.

>
> [..]
>
>>> Can we call this something like 'store_action' or 'remember_action' ?
>>> 'update_list' is quite opaque as name.
>>> Or maybe at least qualify: 'update_stop_list'
>>
>> Yes, I'm not completely happy with the variable name either. How about
>> update_service_list?
>
> Sounds ok.
>
> [..]
>
>>> Will this trigger also when ipa-client-install is run ?
>>> We have a patch on the list to restart sssd via ipa-client-install.
>>> sssd *should* not end in the stop-list though.
>>
>> No, the only services we care about for ipactl are those started by the
>> server itself. I don't think a user would expect that certmonger,
>> messagebus, sssd, etc would stop if they executed ipactl stop.
>
> Yes this is what I am saying, so ipa-client-install is identified as
> 'cli' and not as 'installer' I guess ?
>
> [..]
>

Exactly. When the IPA api is bootstrapped you supply a context name. 
This is generally unique in the IPA universe and different names make 
things do different things.

>> I'll work up a new patch soon.
>
> Thanks,
> Simo.
>