Sebastien Roy wrote:
> David Bustos wrote:
>> Quoth Sebastien Roy on Wed, Oct 10, 2007 at 06:48:02 -0400:
>>> Cathy Zhou wrote:
>>>> 1. Instead of write those "dladm" commands to 
>>>> /var/svc/profile/upgrade, we write it to another script, and run 
>>>> that script at the very beginning of network/physical service.
>>
>> That should work, I think.
> 
> The downside of this one, however, is you now have this upgrade turd in 
> the network/physical method script that is run forever.
> 
>>>> 2. Change the dependency so that instead of system/manifest-import 
>>>> depending on network/physical, we make network/physical depends on 
>>>> system/manifest-import.
>>> Another potential option:
>>>
>>> 3. Create a dlmgmt-upgrade service which disables itself the first 
>>> time it is run.  This service would read the dladm command from the 
>>> custom script described in 1.  The inetd-upgrade service is similar 
>>> to this.
>>
>> I doubt this will work since presumably dlmgmt-upgrade would be
>> installed by droping a manifest into /var/svc/manifest.  In that case
>> the service won't be created until manifest-import anyway.
>>
>> You could pre-import dlmgmt-upgrade's manifest into the target
>> repository with SVCCFG_REPOSITORY.  That's private and in theory can
>> break at any time, though.
> 
Note that dlmgmt-upgrade is not the only manifest we need to import before 
system/manifest-import. There is also a dlmgmtd manifest (the manifest for 
the datalink-management service to start the dlmgmtd daemon).

I talked to David Powell offline and was told that in that case, the only 
option left is to pre-import the service in the package script.

But for the datalink upgrade, as pre-import the manifest during upgrade is 
not generally a good idea, I think we might want to choose option 1.

Thanks
- Cathy

[ The discussion between David and I is forwarded... ]

> I'm not sure I understand.  There has to be a way to introduce a new 
> service to SMF which can exist anywhere in the dependency tree. 
> Otherwise, two reboots would be required after having upgraded a system 
> to a version of Solaris containing such a service.  From what I read 
> above, I'm gathering that because manifest-import starts after 
> network-physical, that we can't introduce a new service which needs to 
> start before network-physical?  I'm confused.
> 
> -Seb

-------------- next part --------------
An embedded message was scrubbed...
From: David Powell <[email protected]>
Subject: Re: [clearview-discuss] code review request for the Clearview UV
        component
Date: Wed, 10 Oct 2007 17:01:37 -0700
Size: 4794
URL: 
<http://mail.opensolaris.org/pipermail/clearview-discuss/attachments/20071011/1d0c0e80/attachment.nws>

Reply via email to