Hi Mary,

On 03/11/09 20:01, mary ding wrote:
> Jan:
>
> Thanks for doing this.  BTW, I just close bugid 7293 as a duplicate of 
> bugid 6612.
>
> We should include SUNWima as well for sparc and this is from bug 7293:
>
> The current AI image doesn't contain
>
> /usr/lib/libsun_ima.so.1
> and
> /usr/lib/64/libsun_ima.so.1
>
> and cause the 'iscsiadm' unable to work.
>
> iscsiadm is the primary utility to configure iSCSI device in 
> OpenSolaris so
> that this should be fixed to make the future support to iSCSI possible in
> OpenSolaris AI.
>
> If SUNWima is not included in this build, can this be fixed.

After talking with people from iSCSI team who encountered this problem,
it was realized that iscsiadm always requires libsun_ima.so.1 in order to
work. Based on this, we agreed that it might be better if SUNWiscsi
depends on SUNWima - in that case it would be pulled automatically
into AI image without need to explicitly specify it in AI DC manifest.

Based on this, I have reopened 7293 and moved into pkg/importer
category, so that this dependency is introduced in IPS dependency chain.

>
> As for myself, so far I had not run across the missing drivers issues 
> on the sparc and x86 machines that I used for testing.

Thank you for verifying - with respect to this I think that
only regression tests are to be done for now.

>   I will also check on other small memory configuration machines like 
> 768 MB for x86 if possible.

Thanks a lot for helping me out with this !
Jan

>
>
>
>
>
> jan damborsky wrote:
>> I have been working on fix for bug 6612
>> (AI image missing some drivers)
>> and now I am in point when I think it would be useful
>> to share where I stand right now and ask people if
>> they might help me to clarify some points.
>>
>> I have done the research what should be included in AI
>> image following Mary's and David's recommendations
>> (please see bug report and preliminary webrev [1]
>> for details) and put together testing AI images based
>> on build 109 (http://ipkg.sfbay/dev IPS repo was used)
>> containing install packages built from following changeset:
>>
>> $ hg tip
>> changeset:   469:8e3825cb3e62
>> tag:         tip
>> user:        Jean McCormack <Jean.McCormack at Sun.COM>
>> date:        Tue Mar 10 07:51:17 2009 -0600
>> summary:     4488 Apache web services are started outside of SMF.
>>
>> Now I am thinking how those should be tested. Looking at bug report
>> the original problem reported by Sean was that AI failed on v480R
>> with hme NIC.
>>
>>
>> Mary, Sean,
>>
>> I would like to ask, are you aware of any other scenarios (HW
>> configuration) where AI failed because of lack of particular
>> device drivers ?
>>
>> If those scenarios exist, could I please ask you to help me to
>> either test or provide access to machines in question or those
>> with HW configuration similar enough (e.g. the same NIC hw, ...),
>> so that I could verify that fix works correctly ?
>> Thank you in advance.
>>
>> I have done regression tests on T1000 (8GB RWM) for Sparc and
>> w2100z (2GB RWM) and Virtual box (800 MB RWM) for x86.
>>
>> Sparc installation went fine.
>>
>> The x86 one is suspicious (I have verified that for me
>> the behavior is the same if I use official RE 109 x86 AI
>> image pulled from nana).
>>
>> The installation itself went fine, but I have noticed following
>> message in install_log file:
>>
>> $ cat /tmp/install_log (full is attached)
>> ...
>> <ICT Mar 11 14:41:00> bootadm update-archive output: 
>> /a//boot/solaris/bin/create_ramdisk: line 509: /usr/sbin/df: cannot 
>> execute [I/O error]
>> ...
>>
>> I have restarted the installation and verified /usr/sbin/df works
>> correctly at the beginning of the installation process. After it
>> finished I can't access it anymore:
>>
>> $ cp /usr/sbin/df /tmp/
>> /tmp//df: I/O error
>> $ file /usr/sbin/df
>> /usr/sbin/df:   cannot read: I/O error
>>
>> There seem to be more files affected:
>>
>> $ find /usr/sbin/ -type f | xargs file | grep cannot
>> /usr/sbin/cron: cannot read: I/O error
>> /usr/sbin/devinfo:      cannot read: I/O error
>> /usr/sbin/df:   cannot read: I/O error
>> /usr/sbin/dfmounts:     cannot read: I/O error
>> /usr/sbin/dfshares:     cannot read: I/O error
>>
>> My question if is this known issue, or if somebody else
>> noticed this - it happens only on x86 for me, I haven't
>> seen this on T1000 Sparc machine.
>>
>> If anyone would like to take a look, I can provide
>> access to x2100z machine suffering from this issue.
>>
>> Thank you very much,
>> Jan
>>
>> -------------------------------------------------------
>>
>> Source of images containing fix for 6612 (available only internally)
>>
>> x86:
>> /net/10.18.138.30/rpool/dc/ai/ai_109_6612/media/automated_installer_image.iso
>>  
>>
>>
>> sparc:
>> /net/tia.czech/data/dc/ai/ai_109_6612/media/automated_installer_image_sparc.iso
>>  
>>
>>
>> preliminary webrev:
>> http://cr.opensolaris.org/~dambi/bug-6612/
>>
>


Reply via email to