Hi Sarah,

when I was doing initial investigation, it seemed to me that
at the moment useradd fails the system is left in inconsistent
state because of steps which happened before useradd was invoked.
This is the reason, why I thought we would need to clean up
the target before installation is restarted after useradd failure.

But then I found out that target instantiation (TI) and transfer (TM)
phases are actually started after useradd finishes and that the
inconsistency was caused by the fact that orchestrator proceeded
with installation even if useradd failed.

It means that if orchestrator returns with failure immediately
when it finds out that useradd wasn't successful and doesn't start
TI nor TM, no actual changes are done to the target, so no cleanup
needs to be done and install can be restarted successfully.

So this fix doesn't address the case when target is in inconsistent
state before installer is invoked - for example if it would fail
for some reason during TI or TM phase and thus target would be
already instantiated and some bits already transfered.

There is still possibility (not addressed by this bug) that
for some reason installer crashes during TI or TM and then
the system might be left in state which would prevent user
from restarting installer successfully - the question might
be if there are valid scenarios when restarting installer
would help to solve the underlying issue in these cases.
Do you think that these possibilities should be more
investigated ?

Thank you very much for the review,
Jan


Sarah Jelinek wrote:
> Hi Jan,
>
> This makes sense to cleanup if the useradd doesn't succeed. but how does 
> this fix cleanup the other issues noted in the bug? Or is this fix 
> intended to do that?
>
> sarah
> ***
> jan damborsky wrote:
>> Hi Sarah, Sundar,
>>
>> could I please ask you to review changes for
>> following bug ?
>>
>> 88 installer goes nuts when useradd fails
>> http://defect.opensolaris.org/bz/show_bug.cgi?id=88
>>
>> Webrev is available at
>> http://cr.opensolaris.org/~dambi/bug-88/
>>
>> Thank you very much,
>> Jan
>>
>>
>>
> _______________________________________________
> caiman-discuss mailing list
> caiman-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss


Reply via email to