mario heimel wrote: > hello, > > until now we have patched our solaris 10 systems (with many zones on all of > them) with scripts, which run patchadd for every patch and 118833-36 was the > last patch in the patchcluster, because of its "you must reboot now code" > anybody knows. > > after reading about pdo magic in this forum i have tried this with the > patchcluster and KU 118833-36.
actually running patchadd -M . in a dor containing 118833-36 should have told you that 118833-36 needs to be installed on it's own, no patches should have installed. I can test this, but from this it appears that it did install patches, which will certainly fail. I always recommend running patchadd -a -M . in the patch dir first, -a is undocumented ( I have an RFE open to get it moved to supported interface ) -a does a dryrun, boots any zones and verifies dependencies in both global and non global zones. It can find certain types of problems up front and allow you to correct them, especially in a zones environment. > > with "patchadd -M `pwd`" pdo magic all patches were installed first in the > global zone and after that in all non-global zones. when 118833-36 is > installed in the global zone, patchadd failed to install the patches in the > non-global zones. > > you get a completely patched global zone but all non-global zones are not > patched. i have tried to install the patchcluster in the non-global zones to > get all in sync but many patches are global-zones-only patches and so there > is no documented way to install the patches with "patchadd" in the non-global > zones. > > after the test i have removed the patchcluster and remain with the old > "patchadd for every patch", which is slower but safer with the beast > 118833-36. > what about 120012-14, it is also a beast ? :) hopefully not as bad ( cannot be installed via patchadd -M ) but does not require a reboot, one can continue to install patches, manually via patchadd. the reason for the enforced reboot in 118833-36 was due to the volumne of change delivered, some of which was not compatible with the running the running system, this required the use of lofs mounts so that the system would not see the "new" files untill post a reboot, we then had to enforce a reboot such that further patchadd operation could continue safely. in 120011 patchadd behaves differently when installign the patch, so the enforced reboot is not required. In general I would recommend the large KU's such as 118833/120011 be installed in single user and in the case of 120011 that a reboot is done once patchadd operations have completed for all patches. In general my recommendation is to use Live Upgrade for patching in Solaris 10 where possible. Thre are two obvious benefits 1 only downtime is time to reboot between the boot partitions 2 if problems are encountered, then it is easy to reverse back to the original boot environment ( the unpatched one ) Enda > > thanks > mario > > > This message posted from opensolaris.org > _______________________________________________ > install-discuss mailing list > install-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/install-discuss
