[OpenIndiana-discuss] New pkg(5) issues

Wed Jun 17 07:00:51 UTC 2015

On 06/16/15 06:04 PM, Jim Klimov wrote:
> 16 июня 2015 г. 15:09:09 CEST, Predrag Zecevic <predrag.zecevic at 2e-systems.com> пишет:
>> Am 16.06.2015 um 14:32 schrieb Jim Klimov:
>>>
>>>   Hello again,
>>>
>>>   Now that I've got a newly upgraded Hipster after all, attempts to
>> update it by running pkg fail due to:
>>>
>>>   ===
>>>
>>>   Linked image publisher checkpkg: update failed (linked image
>> exception(s)):
>>>
>>>   A 'pubcheck-linked' operation failed for child 'zone:mylocalzone'
>> with an unexpected
>>>   return value of 99 and generated the following output:
>>>   Traceback (most recent call last):
>>>     File "/usr/bin/pkg", line 5423, in handle_errors
>>>       __ret = func(*args, **kwargs)
>>>     File "/usr/bin/pkg", line 1815, in __dispatch
>>>       rv = op_func(op, _api_inst, **pwargs)
>>>     File "/usr/bin/pkg", line 4027, in pubcheck_linked
>>>       api_inst.linked_publisher_check()
>>>     File "/usr/lib/python2.6/vendor-packages/pkg/client/api.py", line
>> 206, in wrapper
>>>       return f(instance, *fargs, **f_kwargs)
>>>     File "/usr/lib/python2.6/vendor-packages/pkg/client/api.py", line
>> 1628, in linked_publisher_check
>>>       self.__linked_pubcheck()
>>>     File "/usr/lib/python2.6/vendor-packages/pkg/client/api.py", line
>> 1607, in __linked_pubcheck
>>>       self._img.linked.pubcheck()
>>>     File
>> "/usr/lib/python2.6/vendor-packages/pkg/client/linkedimage/common.py",
>> line 1050, in pubcheck
>>>       linked_pub_error=(pubs, ppubs))
>>>   PlanCreationException:
>>>   Invalid child image publisher configuration.  Child image publisher
>>>   configuration must be a superset of the parent image publisher
>> configuration.
>>>   Please update the child publisher configuration to match the parent.
>> If the
>>>   child image is a zone this can be done automatically by detaching
>> and
>>>   attaching the zone.
>>>
>>>   The parent image has the following enabled publishers:
>>>       PUBLISHER 0: openindiana.org
>>>       PUBLISHER 1: opensolaris.org
>>>
>>>   The child image has the following enabled publishers:
>>>       PUBLISHER 0: on-nightly
>>>       PUBLISHER 1: openindiana.org (non-sticky)
>>>       PUBLISHER 2: opensolaris.org
>>>
>>>
>>>   pkg: This is an internal error in pkg(5) version 2a5f6ab.  Please
>> log a
>>>   Service Request about this issue including the information above and
>> this
>>>   message.
>>>
>>>   ===
>>>
>>>
>>>
>>>
>>>   It does not very much seem true that the child publisher
>> configuration
>>>   is NOT a superset of the parent's one (except for that sticky
>> remark).
>>>
>>>   Oh, and the publishers were different (2014.1 remained in the local
>> zone).
>>>
>>>   While there was little problem removing the non-sticky note in this
>> zone,
>>>   as well as updating the publisher definition, it did not help - same
>> error.
>>>
>>>   Detaching and attaching (without and with "-u") did not succeed and
>> left
>>>   the zone un-mountable so I had to rollback and fix it up.
>>>
>>>   Subsequent upgrade was to "pkg -R /zones/mylocalzone/root/ update"
>>from GZ...
>>>   and it did not solve the issue, either.
>>>
>>>   Only removing the "on-nightly" in the local zone (so publisher lists
>> are now
>>>   identical - which defeats the purpose of building/testing stuff in
>> non-global
>>>   zones) allowed "pkg" to consider an upgrade strategy.
>>>
>>>   I wonder if pkg(5) worked as intended here, or there is a mistake in
>> it indeed?
>>>   Perhaps if linked-zone operation is a no-go, it should revert to old
>> algorithm
>>>   automatically? Or are linked images the new default/requirement?
>>>
>>>   Thanks,
>>>   // Jim Klimov
>>
>> Hi Jim,
>>
>> I also had similar issue: from 2 zones, I was able to (detach/pkg
>> update/reboot/attach -u) ONLY /hipster based zone
>>
>> Another one (based on /dev repository) failed to be handled properly
>> and
>> now is dead (that was probably my mistake during attempts to recover
>>from situation - unfortunatelly I made no notes).
>>
>> Just FYI.
>>
>> In my opinion this error should not happen (maybe some change from
>> Oracle lander untested?).
>>
>> Best regards.
>>> _______________________________________________
>>> openindiana-discuss mailing list
>>> openindiana-discuss at openindiana.org
>>> http://openindiana.org/mailman/listinfo/openindiana-discuss
>>>
>
> Can you at least mount your 'broken' zone (e.g. 'zoneadm mount' or better 'zoneadm ready'? Then you can use 'pkg -R' from the GZ to manipulate IPS metadata and installed bits in the (whole-root) local zone.
>
> I *guess* you can then drive it through the same series of publisher and package updates as we did during Hipster evolution from /dev.
>
> If your zone died mid-way through 'zoneadm attach', you may have to revert to a previous snapshot or 'zbe-NN' clone and fix up attributes (zoned, canmount, mountpoint) of the resulting '.../ROOT/zbe' dataset and/or its parents (ROOT or the ... above).
>
> Good luck!

Hi Jim,

I have fixed problem(s):

a) mountpoint was changed from legacy to some directory (probably due failure of 'zoneadm -z solaruim attach -u' command), reverted 
back to 'legacy' for rpool/Zones/solarium/ROOT

b) after a) was fixed, I was able to boot zone and login into, but all publisher set-up was replaced with one from global zone (so, 
instead of /dev was /hipster, etc). I have reverted that and ran 'pkg update' which has fixed back everything.

Many thanks for tips.
Best regards.
> --
> Typos courtesy of K-9 Mail on my Samsung Android
>

-- 
Predrag Zečević
Technical Support Analyst
2e Systems GmbH

Telephone: +49 6196 9505 815, Facsimile: +49 6196 9505 894
Mobile:    +49  174 3109 288,     Skype: predrag.zecevic
E-mail:    predrag.zecevic at 2e-systems.com

Headquarter:          2e Systems GmbH, Königsteiner Str. 87,
                       65812 Bad Soden am Taunus, Germany
Company registration: Amtsgericht Königstein (Germany), HRB 7303
Managing director:    Phil Douglas

http://www.2e-systems.com/ - Making your business fly!