Loading ...
Sorry, an error occurred while loading the content.

Re: New Solaris 11 (11.1) user questions

Expand Messages
  • David Comay
    ... Although I m not quite sure which interface you re referring to, I suspect it may be the ones having to do with inheritance of share properties. There
    Message 1 of 30 , Jan 18, 2013
    View Source
    • 0 Attachment
      > However, the CIFS change introduced in S11.1 are not enterprise style.
      > They're consumer-linux-distro-of-the-week-style, at best. They're
      > breaking an interface and preventing use of Best Practice upgrades for
      > sysadmins, without bothering to warn them.

      Although I'm not quite sure which interface you're referring to, I
      suspect it may be the ones having to do with inheritance of "share"
      properties. There was a change that took place between Solaris 11
      Express and Solaris 11 FCS that unfortunately introduced some serious
      issues. In order to address those issues, I believe we needed to turn
      off the inheritance of these particular properties but it was noted
      that this was a functional regression from the feature in Solaris 11
      Express. We did not want to remove the new feature that had come into
      Express but at the same time, we definitely did not want to ship a
      subsystem with some important outstanding issues.

      In Solaris 11.1, this feature was reintroduced and the original bugs
      that caused the serious issues were resolved but unfortunately that
      did lead to an incompatibility although it did restore the Express
      functionality.

      In general, we are extremely careful in maintaining compatibility as I
      think you know. I don't know the specific details of whether or not
      this incompatibility was called out in the release notes but if it
      wasn't, my apologies. We should have noted it.
    • Matthias Pfützner
      Just as an intro: I no longer work for Oracle/Sun. ... One of the major difference between S10 and S11 is the packaging mechanism. As such, also the concept
      Message 2 of 30 , Jan 19, 2013
      View Source
      • 0 Attachment
        Just as an intro: I no longer work for Oracle/Sun.

        You (John D Groenveld) wrote:
        > In message <50F99913.4000806@...>, Alan Coopersmith writes:
        > >We no longer make SRU's for Solaris 11 - customers with support should upgrade
        > >to the Solaris 11.1 SRU's.
        >
        > Thank you for the heads-up, I hadn't realized Chuck Rozwat and
        > company had EOLd 11.0.

        One of the major difference between S10 and S11 is the packaging mechanism. As
        such, also the "concept of patches" changed. There no longer are patches!
        There are "versions of packages", and as such, there are "dependencies between
        these packages".

        So, yes, S11 is still supported! BUT: The individual packages might have
        different version numbers...

        > John
        > groenveld@...

        Matthias
        --
        Matthias Pfützner | Tel.: +49 171 4735638 | Things have changed in
        Waldstrasse 6 | mailto:Matthias@... | the past two decades.
        D-61440 Oberursel | AIM: pfuetz, ICQ: 300967487 |
        Germany | http://www.pfuetzner.de/matthias/ | Bill Gates, 1995
      • John D Groenveld
        In message , Matthias =?iso-8859-1?Q?Pf=FC ... You and a lot of others. ... The move from
        Message 3 of 30 , Jan 19, 2013
        View Source
        • 0 Attachment
          In message <20130119142951.GB14554@...>, Matthias =?iso-8859-1?Q?Pf=FC
          tzner?= writes:
          >Just as an intro: I no longer work for Oracle/Sun.

          You and a lot of others.

          >One of the major difference between S10 and S11 is the packaging mechanism. As
          >such, also the "concept of patches" changed. There no longer are patches!
          >There are "versions of packages", and as such, there are "dependencies between
          >these packages".
          >
          >So, yes, S11 is still supported! BUT: The individual packages might have
          >different version numbers...

          The move from pkg://solaris/entire@0.5.11,5.11-0.175.0.13.0.4.0:20121106T194623Z
          to pkg://solaris/entire@0.5.11,5.11-0.175.1.3.0.4.1:20130115T215210Z
          may not be as big as Sun Solaris 2.5 to 2.5.1 or Oracle RDBMS 11.1
          to 11.2, but it is still significant and warrants an EOL notice.

          John
          groenveld@...
        • Bob Netherton
          ... That is like saying that Sun (and later Oracle) should have issued an EOL notice for Solaris 10 update 2. John, you are way over-thinking the wrong part of
          Message 4 of 30 , Jan 19, 2013
          View Source
          • 0 Attachment
            On 01/19/13 08:41 AM, John D Groenveld wrote:
            >
            >
            > >
            > >So, yes, S11 is still supported! BUT: The individual packages might have
            > >different version numbers...
            >
            > The move from pkg://solaris/entire@0.5.11,5.11-0.175.0.13.0.4.0:20121106T194623Z
            > to pkg://solaris/entire@0.5.11,5.11-0.175.1.3.0.4.1:20130115T215210Z
            > may not be as big as Sun Solaris 2.5 to 2.5.1 or Oracle RDBMS 11.1
            > to 11.2, but it is still significant and warrants an EOL notice.
            >

            That is like saying that Sun (and later Oracle) should have issued an EOL notice
            for Solaris 10 update 2.

            John, you are way over-thinking the wrong part of this and missing the big part
            (insert forest/tree analogy). This isn't all that different than Solaris 10
            over the last half decade, except that the packaging system is trying to keep
            you from turning your systems into an unmanageable mess.

            Using kernel patches as the example - in Solaris 10, you would get kernel
            patches for an update for a period of time. When a new update was released, a
            new kernel patch train started and you would no longer get updates to the old
            patch. Apply this same logic to the rest of the Solaris, either directly (not
            thinking of any good examples here) or by dependency on the new kernel patch
            number sequence.

            This is what allowed users to either patch to a compatible update level or
            upgrade to that same level. The difference would generally be new things not
            delivered in a patch, obsolete package removals (staroffice as an example) or
            some other reorganization that might happen along the way.

            That led to things like....

            Can you patch your way to ZFS boot while running s10u5, or how about running u8
            on a T4 ?

            S11.1 is a new kernel and a new set of functionality, as before (like moving
            from s10u2 to s10u3). If you want to dive into the facets of individual
            packages and break the clarity of the entire consolidation dependency, you can
            probably come really close to what you had before. I would recommend against it
            though as it defeats one of the best features of IPS, insuring that dependencies
            are really followed (including the ones I missed in the documentation) - most of
            them are there for a reason.


            As for David's comment earlier, although the history of the share property
            inheritance in s11ga being a regression that was corrected in s11.1, it is
            covered in the What's New section of the 11.1 documentation.


            Bob
          • Alan Coopersmith
            ... They haven t, it s just that the support updates now include all of Solaris 11.1 as well, so the SRU s are now numbered 11.1.x. This isn t really a whole
            Message 5 of 30 , Jan 19, 2013
            View Source
            • 0 Attachment
              On 01/18/13 10:58 AM, John D Groenveld wrote:
              > In message <50F99913.4000806@...>, Alan Coopersmith writes:
              >> We no longer make SRU's for Solaris 11 - customers with support should upgrade
              >> to the Solaris 11.1 SRU's.
              >
              > Thank you for the heads-up, I hadn't realized Chuck Rozwat and
              > company had EOLd 11.0.

              They haven't, it's just that the support updates now include all of Solaris 11.1
              as well, so the SRU's are now numbered 11.1.x.

              This isn't really a whole lot different than Solaris 10 - if you need a kernel
              fix for S10U6, you'll get a kernel patch that includes all the S10U7-U10 changes
              as well, and the recommended patch set delivers half the update release with it.

              It is a lot more explicit, and a lot better tested, since we put a lot more
              effort into testing 11.1 and each SRU than we do into trying to determine
              every possible hybrid combination of S10Ux + patches that customers could
              possibly end up on.

              --
              -Alan Coopersmith- alan.coopersmith@...
              Oracle Solaris Engineering - http://blogs.oracle.com/alanc
            • Laurent Blume
              ... Do you mean that ISV have been warned, and that an application certified for Solaris 11.0 will automatically be so under 11.1? Because that s the kind of
              Message 6 of 30 , Jan 19, 2013
              View Source
              • 0 Attachment
                On 01/19/13 16:50, Bob Netherton wrote:
                > That is like saying that Sun (and later Oracle) should have issued an EOL notice
                > for Solaris 10 update 2.

                Do you mean that ISV have been warned, and that an application certified
                for Solaris 11.0 will automatically be so under 11.1? Because that's the
                kind of game plenty of them have been doing, certifying for a given
                minor version.
                Or are people going to be stuck on 11.0 to be able to get support for
                their apps?

                > As for David's comment earlier, although the history of the share property
                > inheritance in s11ga being a regression that was corrected in s11.1, it is
                > covered in the What's New section of the 11.1 documentation.

                I'm starting to think I'm dense on this one, because neither you nor
                David appear to have seen what I saw.
                I didn't see anything in the Release Notes that said that existing CIFS
                shares would stop working after updating to 11.1, nor that to get them
                working again, you would be forced to upgrade the pool version (and lose
                the ability to use older BE's).
                Did I miss it? What did I do wrong?
                Please note that my existing CIFS shares were on a v28/v4 dataset.
                However, I did check that on a v33/v5, the sharesmb property was the same.
                So, two questions, and I hope those are clear:
                - is there a way to keep CIFS shares made in 11,0 working on S11.1
                without upgrading the pool or FS?
                - if upgrading is needed, is there some kind of magic that converts
                those properties to the new ones in v34/v6?

                thanks,

                Laurent
              • palowoda
                ... Well I do agree IPS is a great improvement in this area. I m still scratching my head up near the top of the release notes (Update Considerations): Note -
                Message 7 of 30 , Jan 20, 2013
                View Source
                • 0 Attachment
                  --- In solarisx86@yahoogroups.com, Bob Netherton wrote:

                  > S11.1 is a new kernel and a new set of functionality, as before (like moving
                  > from s10u2 to s10u3). If you want to dive into the facets of individual
                  > packages and break the clarity of the entire consolidation dependency, you can
                  > probably come really close to what you had before. I would recommend against it
                  > though as it defeats one of the best features of IPS, insuring that dependencies
                  > are really followed (including the ones I missed in the documentation) - most of
                  > them are there for a reason.
                  >

                  Well I do agree IPS is a great improvement in this area. I'm still scratching my head up near the top of the release notes (Update Considerations):

                  Note - The update process fails if you are running Oracle Solaris 11 SRU 12 or later with BIND or Fetchmail installed. To overcome this issue, see the workaround for bugs 7203326 and 7200467.

                  Wouldn't an upgrade test with all packages installed catch this kind of bug?

                  ---Bob
                • Ian Collins
                  ... That s why anyone who wasn t desperate waited for the first 11.1 SRU before upgrading! -- Ian.
                  Message 8 of 30 , Jan 20, 2013
                  View Source
                  • 0 Attachment
                    palowoda wrote:
                    >
                    > --- In solarisx86@yahoogroups.com, Bob Netherton wrote:
                    >
                    >> S11.1 is a new kernel and a new set of functionality, as before (like moving
                    >> from s10u2 to s10u3). If you want to dive into the facets of individual
                    >> packages and break the clarity of the entire consolidation dependency, you can
                    >> probably come really close to what you had before. I would recommend against it
                    >> though as it defeats one of the best features of IPS, insuring that dependencies
                    >> are really followed (including the ones I missed in the documentation) - most of
                    >> them are there for a reason.
                    >>
                    > Well I do agree IPS is a great improvement in this area. I'm still scratching my head up near the top of the release notes (Update Considerations):
                    >
                    > Note - The update process fails if you are running Oracle Solaris 11 SRU 12 or later with BIND or Fetchmail installed. To overcome this issue, see the workaround for bugs 7203326 and 7200467.
                    >
                    > Wouldn't an upgrade test with all packages installed catch this kind of bug?

                    That's why anyone who wasn't desperate waited for the first 11.1 SRU
                    before upgrading!

                    --
                    Ian.
                  • John D Groenveld
                    ... What catch-22 s await when Solaris 11.2 ships and 11.1 is EOLd? Nothing alarming in the EOF notice:
                    Message 9 of 30 , Jan 20, 2013
                    View Source
                    • 0 Attachment
                      In message <50FBC224.4070804@...>, Ian Collins writes:
                      >That's why anyone who wasn't desperate waited for the first 11.1 SRU
                      >before upgrading!

                      What catch-22's await when Solaris 11.2 ships and 11.1 is EOLd?
                      Nothing alarming in the EOF notice:
                      <URL:http://www.oracle.com/technetwork/systems/end-of-notices/eonsolaris11-392732.html>

                      What are Oracle's policies regarding Solaris 11 lifecycle and
                      interface stability?
                      John
                      groenveld@...
                    • palowoda
                      ... If you want read a good reason for the SRU read the thread: https://forums.oracle.com/forums/thread.jspa?threadID=2458513&tstart=0 But SRU s are not what
                      Message 10 of 30 , Jan 20, 2013
                      View Source
                      • 0 Attachment
                        --- In solarisx86@yahoogroups.com, Ian Collins wrote:
                        >
                        > palowoda wrote:
                        > >
                        > > --- In solarisx86@yahoogroups.com, Bob Netherton wrote:
                        > >
                        > >> S11.1 is a new kernel and a new set of functionality, as before (like moving
                        > >> from s10u2 to s10u3). If you want to dive into the facets of individual
                        > >> packages and break the clarity of the entire consolidation dependency, you can
                        > >> probably come really close to what you had before. I would recommend against it
                        > >> though as it defeats one of the best features of IPS, insuring that dependencies
                        > >> are really followed (including the ones I missed in the documentation) - most of
                        > >> them are there for a reason.
                        > >>
                        > > Well I do agree IPS is a great improvement in this area. I'm still scratching my head up near the top of the release notes (Update Considerations):
                        > >
                        > > Note - The update process fails if you are running Oracle Solaris 11 SRU 12 or later with BIND or Fetchmail installed. To overcome this issue, see the workaround for bugs 7203326 and 7200467.
                        > >
                        > > Wouldn't an upgrade test with all packages installed catch this kind of bug?
                        >
                        > That's why anyone who wasn't desperate waited for the first 11.1 SRU
                        > before upgrading!
                        >

                        If you want read a good reason for the SRU read the thread:

                        https://forums.oracle.com/forums/thread.jspa?threadID=2458513&tstart=0

                        But SRU's are not what really bothers me. My cats bother me the most these days. But maybe a better explanation of why that note kind of makes me itch.

                        Many years ago I was on the WOS committee at Sun for 2.4 and 2.5. As part of it three of us Tim Marsland, Kevin McCallister (if I remember right) and me had the responsibility of weekly reviews of the internal release notes for each build and the final release notes. Maybe something along the line of what David is doing now. But I don't claim it's anything like it was now as it was back than.

                        As you know Solaris 10 and prior you would be updating min, entire installations etc. To come up with a release note that said you had to make changes prior if you had installed the entire system would most likely been rejected and had to been fixed before released. Not put it as a release note.

                        Now fast forward to IPS and why it was designed. To make a better quality software framework that is easier and faster to test. Sun PIT testing and now Oracle has some of the best testing I've ever seen. So what was/is the point of telling a workaround and spending millions on a new install framework and testing just to tell the customer in the release note "Here is the problem your going to have and how to fix it before hand". I don't know the logistics behind the reasoning timing, management dissension, testing etc. I do kind of wonder about if BIND is installed on your system the upgrade is going to fail if you didn't catch it in the release notes. Solaris is considered a server OS.

                        I could be way off base here. But why should be there a special "Note" about these two packages before you upgrade?

                        ---Bob
                      • Laurent Blume
                        ... Forget about it. The CIFS server was having some fits AGAIN. Countless hours refusing my password with no explanation, me checking and re-reading the
                        Message 11 of 30 , Jan 20, 2013
                        View Source
                        • 0 Attachment
                          On 20/01/2013 01:25, Laurent Blume wrote:
                          > So, two questions, and I hope those are clear:
                          > - is there a way to keep CIFS shares made in 11,0 working on S11.1
                          > without upgrading the pool or FS?
                          > - if upgrading is needed, is there some kind of magic that converts
                          > those properties to the new ones in v34/v6?
                          >
                          > thanks,

                          Forget about it. The CIFS server was having some fits AGAIN. Countless
                          hours refusing my password with no explanation, me checking and
                          re-reading the documentation until I gave up. It now decided to accept
                          it. After a few reboots. Or maybe it only works on weekends. Or
                          depending on the moon. Or when it snows.

                          I apologize for comparing it to Linux. This is windows95-y software.

                          I'm through with that timewaster, to be polite. Back to Samba, software
                          that works. Less ranting, more use :-)

                          Laurent
                        • Alan Coopersmith
                          ... Testing can t find issues that don t exist yet. The problem was caused by allowing security fixes into the SRU repo that delivered newer versions of
                          Message 12 of 30 , Jan 20, 2013
                          View Source
                          • 0 Attachment
                            On 01/20/13 12:41 AM, palowoda wrote:
                            > Note - The update process fails if you are running Oracle Solaris 11 SRU 12 or later with BIND or Fetchmail installed. To overcome this issue, see the workaround for bugs 7203326 and 7200467.
                            >
                            > Wouldn't an upgrade test with all packages installed catch this kind of bug?

                            Testing can't find issues that don't exist yet. The problem was caused by
                            allowing security fixes into the SRU repo that delivered newer versions of
                            those packages after the 11.1 release was done.

                            https://blogs.oracle.com/Solaris11Life/entry/solaris_11_process_enhancement_no
                            http://www.oracle.com/technetwork/articles/servers-storage-admin/howto-update-11dot1-ips-1866781.html

                            --
                            -Alan Coopersmith- alan.coopersmith@...
                            Oracle Solaris Engineering - http://blogs.oracle.com/alanc
                          • palowoda
                            ... Ughh. I guess that testing and release policy didn t work out well. On the plus side at least IPS hangs and gets the attention to go read the release notes
                            Message 13 of 30 , Jan 20, 2013
                            View Source
                            • 0 Attachment
                              --- In solarisx86@yahoogroups.com, Alan Coopersmith wrote:
                              >
                              > On 01/20/13 12:41 AM, palowoda wrote:
                              > > Note - The update process fails if you are running Oracle Solaris 11 SRU 12 or later with BIND or Fetchmail installed. To overcome this issue, see the workaround for bugs 7203326 and 7200467.
                              > >
                              > > Wouldn't an upgrade test with all packages installed catch this kind of bug?
                              >
                              > Testing can't find issues that don't exist yet. The problem was caused by
                              > allowing security fixes into the SRU repo that delivered newer versions of
                              > those packages after the 11.1 release was done.
                              >

                              Ughh. I guess that testing and release policy didn't work out well. On the plus side at least IPS hangs and gets the attention to go read the release notes for the workaround.

                              ---Bob
                            • Ian Collins
                              ... This one got me: http://www.mail-archive.com/zfs-discuss@opensolaris.org/msg50263.html ... To be fair, it was only paying support customers who were
                              Message 14 of 30 , Jan 20, 2013
                              View Source
                              • 0 Attachment
                                palowoda wrote:

                                > If you want read a good reason for the SRU read the thread:
                                >
                                > https://forums.oracle.com/forums/thread.jspa?threadID=2458513&tstart=0

                                This one got me:

                                http://www.mail-archive.com/zfs-discuss@.../msg50263.html

                                > But SRU's are not what really bothers me. My cats bother me the most these days. But maybe a better explanation of why that note kind of makes me itch.
                                >
                                > Many years ago I was on the WOS committee at Sun for 2.4 and 2.5. As part of it three of us Tim Marsland, Kevin McCallister (if I remember right) and me had the responsibility of weekly reviews of the internal release notes for each build and the final release notes. Maybe something along the line of what David is doing now. But I don't claim it's anything like it was now as it was back than.
                                >
                                > As you know Solaris 10 and prior you would be updating min, entire installations etc. To come up with a release note that said you had to make changes prior if you had installed the entire system would most likely been rejected and had to been fixed before released. Not put it as a release note.
                                >
                                > Now fast forward to IPS and why it was designed. To make a better quality software framework that is easier and faster to test. Sun PIT testing and now Oracle has some of the best testing I've ever seen. So what was/is the point of telling a workaround and spending millions on a new install framework and testing just to tell the customer in the release note "Here is the problem your going to have and how to fix it before hand". I don't know the logistics behind the reasoning timing, management dissension, testing etc. I do kind of wonder about if BIND is installed on your system the upgrade is going to fail if you didn't catch it in the release notes. Solaris is considered a server OS.
                                >
                                > I could be way off base here. But why should be there a special "Note" about these two packages before you upgrade?

                                To be fair, it was only paying support customers who were impacted by
                                this :)

                                I took one look at the update notes here:

                                http://www.oracle.com/technetwork/articles/servers-storage-admin/howto-update-11dot1-ips-1866781.html

                                and waited for the first SRU before upgrading. Then I lost my test
                                system's pool, recovered it and applied the second SRU...

                                Maybe later this month I might upgrade some production systems.

                                --
                                Ian.
                              • John D Groenveld
                                ... I usuaully see timeout failures from pkg.Oracle.COM, then authentication/authorization errors, and then it returns to life within an hour. The workaround
                                Message 15 of 30 , Jan 21, 2013
                                View Source
                                • 0 Attachment
                                  In message <kdhknt+ku4l@...>, "palowoda" writes:
                                  > Ughh. I guess that testing and release policy didn't work out well. On the pl
                                  >us side at least IPS hangs and gets the attention to go read the release notes
                                  > for the workaround.

                                  I usuaully see timeout failures from pkg.Oracle.COM, then
                                  authentication/authorization errors, and then it returns
                                  to life within an hour.

                                  The workaround is to download the ISO image from Chuck Rozwat
                                  and company's support portal and set up a mirror or local
                                  file repo.

                                  John
                                  groenveld@...
                                Your message has been successfully submitted and would be delivered to recipients shortly.