hi All, What patches should be applied for 1.6.5.1 to get it more stable (also the same question for patchless clients)? I''ve found this in a bugzilla comment for clients: https://bugzilla.lustre.org/attachment.cgi?id=18103&action=edit And this at the list: https://bugzilla.lustre.org/show_bug.cgi?id=16496 What else patches are proposed to use? Is there any chance in the future to make these patches collected somehow (for example more frequently source-only mini release , like 1.6.5.x, or a website, or something like that)? Thanks, tamas
On Fri, 2008-10-10 at 22:12 +0200, Papp Tam?s wrote:> hi All,Hi,> What patches should be applied for 1.6.5.1 to get it more stable (also > the same question for patchless clients)?Well, by definition, all of the patches that will be applied to release n are needed by release n-1. For 1.6.5.1 that would be all of the patches we have scheduled to land for 1.6.6. That may sound like a silly answer, but that really is the truth. It''s not quite that simple however. Until a release (let''s call it 1.6.6 for example) has gone through our QA process, there is no assurance that there are not inter-patch defects (more commonly referred to as ripple effects). Otherwise we could just apply a patch and make 1.6.5.2 and apply another and call it 1.6.5.3, etc.> I''ve found this in a bugzilla comment for clients: > > https://bugzilla.lustre.org/attachment.cgi?id=18103&action=editYes, that one can be an important one.> And this at the list: > > https://bugzilla.lustre.org/show_bug.cgi?id=16496Not sure how important this one is, but as I will get to, it''s all relative.> What else patches are proposed to use?We had recently advised one customer recently to apply the following attachments: attachment 18121 attachment 18293 attachment 19123 attachment 18103 In addition to a few others that were to fix problems particular to their installation. But this is the rub. Which patches any site should apply to a release depends on what their use patterns are like and which bugs they are running into as a result. Given the warning above about ripple effects, applying patches to a given release is a careful balance between minimizing risks of ripple effects and getting the fixes for the particular problems that are ailing you. Obviously the more testing you can do of your release +patches build before going live, the more patches you should be able to sustain without rude go-live incidents.> Is there any chance in the future to make these patches collected > somehow (for example more frequently source-only mini release , like > 1.6.5.x, or a website, or something like that)?As for more frequent releases, the problem we have with that is that our QA process is quite involved due to the complexity of Lustre and the array of kernels and architectures we have to test on to call any given release "fully QA certified". As can be witnessed by 1.6.5.1 (and the few 1.6.4.x releases) however, occasionally the need is justified. As for a website, I suppose you are looking for an enumeration of patches that will be applied to release n to get n+1. You can get information that from our bugzilla in the form of the release tracking bugs. For any given release there is a bug tracking all of the bugs that will be applied to release n-1 to get release n. We usually give those tracking bugs an alias of the form wxy[z]-tracking where w, x, y and possible z are the digits from the release. So for 1.6.6 you''d be interested in "166-tracking" which you can put into a "bug number" field in our BZ and you will get bug 14883. All of the bugs in the Depends on: field are the bugs with patches that have gone into 1.6.5 to get 1.6.6. If you wanted to be notified when a new bug has been identified as needing to go into a release you could simply subscribe yourself to the release tracker and you will get notified of updates to that Depends on: field. Hopefully this tells you all you need to know (and more) about our release process.
Brian J. Murrell wrote:> Hopefully this tells you all you need to know (and more) about our > release process. >Thanks, I hope too:) tamas
Hi, When is release date for 1.6.6? Thanks On Sat, Oct 11, 2008 at 4:02 AM, Papp Tamas <tompos at martos.bme.hu> wrote:> Brian J. Murrell wrote: > > Hopefully this tells you all you need to know (and more) about our > > release process. > > > > Thanks, I hope too:) > > tamas > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at lists.lustre.org > http://lists.lustre.org/mailman/listinfo/lustre-discuss >-------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.lustre.org/pipermail/lustre-discuss/attachments/20081011/656dec17/attachment.html
On Sat, 2008-10-11 at 07:41 +0700, sd a wrote:> Hi, > > When is release date for 1.6.6?I think I already answered today that I don''t know when 1.6.6 will be released because for the most part, nobody does. We don''t make release date promises or predictions because we don''t even know the date until the QA cycle for a release candidate is complete in which case the release date is for all intents and purposes that day -- the overhead for getting packages from QA to the SDLC, etc. aside. Each and every release candidate that goes to QA has the potential to be the GA release. It also has the potential to not be. That''s QA doing it''s job for you. :-) b.
Would there be a problem if we have 5 servers running 1.6.6 and the remainder of them running 1.6.5.1? Is there a chance of data corruption? On Fri, Oct 10, 2008 at 10:08 PM, Brian J. Murrell <Brian.Murrell at sun.com> wrote:> On Sat, 2008-10-11 at 07:41 +0700, sd a wrote: >> Hi, >> >> When is release date for 1.6.6? > > I think I already answered today that I don''t know when 1.6.6 will be > released because for the most part, nobody does. > > We don''t make release date promises or predictions because we don''t even > know the date until the QA cycle for a release candidate is complete in > which case the release date is for all intents and purposes that day -- > the overhead for getting packages from QA to the SDLC, etc. aside. > > Each and every release candidate that goes to QA has the potential to be > the GA release. It also has the potential to not be. That''s QA doing > it''s job for you. :-) > > b. > > > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at lists.lustre.org > http://lists.lustre.org/mailman/listinfo/lustre-discuss >
I meant to say, 5 clients running 1.6.6 and the remainder running 1.6.5.1 On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam <magawake at gmail.com> wrote:> Would there be a problem if we have 5 servers running 1.6.6 and the > remainder of them running 1.6.5.1? Is there a chance of data > corruption? > > > > On Fri, Oct 10, 2008 at 10:08 PM, Brian J. Murrell > <Brian.Murrell at sun.com> wrote: >> On Sat, 2008-10-11 at 07:41 +0700, sd a wrote: >>> Hi, >>> >>> When is release date for 1.6.6? >> >> I think I already answered today that I don''t know when 1.6.6 will be >> released because for the most part, nobody does. >> >> We don''t make release date promises or predictions because we don''t even >> know the date until the QA cycle for a release candidate is complete in >> which case the release date is for all intents and purposes that day -- >> the overhead for getting packages from QA to the SDLC, etc. aside. >> >> Each and every release candidate that goes to QA has the potential to be >> the GA release. It also has the potential to not be. That''s QA doing >> it''s job for you. :-) >> >> b. >> >> >> >> _______________________________________________ >> Lustre-discuss mailing list >> Lustre-discuss at lists.lustre.org >> http://lists.lustre.org/mailman/listinfo/lustre-discuss >> >
On Oct 11, 2008 12:15 -0400, Mag Gam wrote:> I meant to say, 5 clients running 1.6.6 and the remainder running 1.6.5.1 > > > On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam <magawake at gmail.com> wrote: > > Would there be a problem if we have 5 servers running 1.6.6 and the > > remainder of them running 1.6.5.1? Is there a chance of data > > corruption?No, we always test 1.x.y against 1.x.y-1 and 1.x-1.latest during our QA process. In general there should never be any chance of data corruption no matter what two releases you try to use together, because if there is (and we know about it, of course) then those versions would explicitly be detected as non-interoperable at connect time. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc.
Thankyou very much. I have done this for 48 hours now and no problem to report. The new rc version is remarkable and the statahead problems aren''t occuring anymore! Thanks again On Sun, Oct 12, 2008 at 2:12 AM, Andreas Dilger <adilger at sun.com> wrote:> On Oct 11, 2008 12:15 -0400, Mag Gam wrote: >> I meant to say, 5 clients running 1.6.6 and the remainder running 1.6.5.1 >> >> >> On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam <magawake at gmail.com> wrote: >> > Would there be a problem if we have 5 servers running 1.6.6 and the >> > remainder of them running 1.6.5.1? Is there a chance of data >> > corruption? > > No, we always test 1.x.y against 1.x.y-1 and 1.x-1.latest during our > QA process. In general there should never be any chance of data corruption > no matter what two releases you try to use together, because if there > is (and we know about it, of course) then those versions would explicitly > be detected as non-interoperable at connect time. > > Cheers, Andreas > -- > Andreas Dilger > Sr. Staff Engineer, Lustre Group > Sun Microsystems of Canada, Inc. > >
I have some spare hardware (20TB of storage and several PE2950 servers) and I would like to use it as a test platform for new lustre version. I noticed that some people talk about testing beta version of 1.6.6. Can some one tell me where could I obtain rc version of 1.6.6 ? Many thanks, Mag Gam wrote: Thankyou very much. I have done this for 48 hours now and no problem to report. The new rc version is remarkable and the statahead problems aren''t occuring anymore! Thanks again On Sun, Oct 12, 2008 at 2:12 AM, Andreas Dilger wrote: On Oct 11, 2008 12:15 -0400, Mag Gam wrote: I meant to say, 5 clients running 1.6.6 and the remainder running 1.6.5.1 On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam wrote: Would there be a problem if we have 5 servers running 1.6.6 and the remainder of them running 1.6.5.1? Is there a chance of data corruption? No, we always test 1.x.y against 1.x.y-1 and 1.x-1.latest during our QA process. In general there should never be any chance of data corruption no matter what two releases you try to use together, because if there is (and we know about it, of course) then those versions would explicitly be detected as non-interoperable at connect time. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc. Lustre-discuss mailing list Lustre-discuss-aLEFhgZF4x6X6Mz3xDxJMA@public.gmane.org http://lists.lustre.org/mailman/listinfo/lustre-discuss _______________________________________________ Lustre-discuss mailing list Lustre-discuss-aLEFhgZF4x6X6Mz3xDxJMA@public.gmane.org http://lists.lustre.org/mailman/listinfo/lustre-discuss
!.6.6 is released and should be available from the down load center. J.D. Wojciech Turek wrote:> I have some spare hardware (20TB of storage and several PE2950 > servers) and I would like to use it as a test platform for new lustre > version. I noticed that some people talk about testing beta version > of 1.6.6. Can some one tell me where could I obtain rc version of 1.6.6 ? > > Many thanks, > > Mag Gam wrote: >> Thankyou very much. I have done this for 48 hours now and no problem >> to report. The new rc version is remarkable and the statahead problems >> aren''t occuring anymore! >> >> Thanks again >> >> >> On Sun, Oct 12, 2008 at 2:12 AM, Andreas Dilger <adilger at sun.com> wrote: >> >>> On Oct 11, 2008 12:15 -0400, Mag Gam wrote: >>> >>>> I meant to say, 5 clients running 1.6.6 and the remainder running 1.6.5.1 >>>> >>>> >>>> On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam <magawake at gmail.com> wrote: >>>> >>>>> Would there be a problem if we have 5 servers running 1.6.6 and the >>>>> remainder of them running 1.6.5.1? Is there a chance of data >>>>> corruption? >>>>> >>> No, we always test 1.x.y against 1.x.y-1 and 1.x-1.latest during our >>> QA process. In general there should never be any chance of data corruption >>> no matter what two releases you try to use together, because if there >>> is (and we know about it, of course) then those versions would explicitly >>> be detected as non-interoperable at connect time. >>> >>> Cheers, Andreas >>> -- >>> Andreas Dilger >>> Sr. Staff Engineer, Lustre Group >>> Sun Microsystems of Canada, Inc. >>> >>> >>> >> _______________________________________________ >> Lustre-discuss mailing list >> Lustre-discuss at lists.lustre.org >> http://lists.lustre.org/mailman/listinfo/lustre-discuss >> > > ------------------------------------------------------------------------ > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at lists.lustre.org > http://lists.lustre.org/mailman/listinfo/lustre-discuss >
Indeed you right! I used to get messages from lustre-announce list about such events but it seems it didn''t work this time. Thanks JD Neumann wrote:> !.6.6 is released and should be available from the down load center. > J.D. > > Wojciech Turek wrote: >> I have some spare hardware (20TB of storage and several PE2950 >> servers) and I would like to use it as a test platform for new lustre >> version. I noticed that some people talk about testing beta version >> of 1.6.6. Can some one tell me where could I obtain rc version of >> 1.6.6 ? >> >> Many thanks, >> >> Mag Gam wrote: >>> Thankyou very much. I have done this for 48 hours now and no problem >>> to report. The new rc version is remarkable and the statahead problems >>> aren''t occuring anymore! >>> >>> Thanks again >>> >>> >>> On Sun, Oct 12, 2008 at 2:12 AM, Andreas Dilger <adilger at sun.com> >>> wrote: >>> >>>> On Oct 11, 2008 12:15 -0400, Mag Gam wrote: >>>> >>>>> I meant to say, 5 clients running 1.6.6 and the remainder running >>>>> 1.6.5.1 >>>>> >>>>> >>>>> On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam <magawake at gmail.com> wrote: >>>>> >>>>>> Would there be a problem if we have 5 servers running 1.6.6 and the >>>>>> remainder of them running 1.6.5.1? Is there a chance of data >>>>>> corruption? >>>>>> >>>> No, we always test 1.x.y against 1.x.y-1 and 1.x-1.latest during our >>>> QA process. In general there should never be any chance of data >>>> corruption >>>> no matter what two releases you try to use together, because if there >>>> is (and we know about it, of course) then those versions would >>>> explicitly >>>> be detected as non-interoperable at connect time. >>>> >>>> Cheers, Andreas >>>> -- >>>> Andreas Dilger >>>> Sr. Staff Engineer, Lustre Group >>>> Sun Microsystems of Canada, Inc. >>>> >>>> >>>> >>> _______________________________________________ >>> Lustre-discuss mailing list >>> Lustre-discuss at lists.lustre.org >>> http://lists.lustre.org/mailman/listinfo/lustre-discuss >>> >> >> ------------------------------------------------------------------------ >> >> _______________________________________________ >> Lustre-discuss mailing list >> Lustre-discuss at lists.lustre.org >> http://lists.lustre.org/mailman/listinfo/lustre-discuss >> >
I will check on this and see why it was not sent. J.D. Wojciech Turek wrote:> Indeed you right! I used to get messages from lustre-announce list > about such events but it seems it didn''t work this time. > > Thanks > > JD Neumann wrote: >> !.6.6 is released and should be available from the down load center. >> J.D. >> >> Wojciech Turek wrote: >>> I have some spare hardware (20TB of storage and several PE2950 >>> servers) and I would like to use it as a test platform for new >>> lustre version. I noticed that some people talk about testing beta >>> version of 1.6.6. Can some one tell me where could I obtain rc >>> version of 1.6.6 ? >>> >>> Many thanks, >>> >>> Mag Gam wrote: >>>> Thankyou very much. I have done this for 48 hours now and no problem >>>> to report. The new rc version is remarkable and the statahead problems >>>> aren''t occuring anymore! >>>> >>>> Thanks again >>>> >>>> >>>> On Sun, Oct 12, 2008 at 2:12 AM, Andreas Dilger <adilger at sun.com> >>>> wrote: >>>> >>>>> On Oct 11, 2008 12:15 -0400, Mag Gam wrote: >>>>> >>>>>> I meant to say, 5 clients running 1.6.6 and the remainder running >>>>>> 1.6.5.1 >>>>>> >>>>>> >>>>>> On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam <magawake at gmail.com> wrote: >>>>>> >>>>>>> Would there be a problem if we have 5 servers running 1.6.6 and the >>>>>>> remainder of them running 1.6.5.1? Is there a chance of data >>>>>>> corruption? >>>>>>> >>>>> No, we always test 1.x.y against 1.x.y-1 and 1.x-1.latest during our >>>>> QA process. In general there should never be any chance of data >>>>> corruption >>>>> no matter what two releases you try to use together, because if there >>>>> is (and we know about it, of course) then those versions would >>>>> explicitly >>>>> be detected as non-interoperable at connect time. >>>>> >>>>> Cheers, Andreas >>>>> -- >>>>> Andreas Dilger >>>>> Sr. Staff Engineer, Lustre Group >>>>> Sun Microsystems of Canada, Inc. >>>>> >>>>> >>>>> >>>> _______________________________________________ >>>> Lustre-discuss mailing list >>>> Lustre-discuss at lists.lustre.org >>>> http://lists.lustre.org/mailman/listinfo/lustre-discuss >>>> >>> >>> ------------------------------------------------------------------------ >>> >>> >>> _______________________________________________ >>> Lustre-discuss mailing list >>> Lustre-discuss at lists.lustre.org >>> http://lists.lustre.org/mailman/listinfo/lustre-discuss >>> >> >
JD I think that the announcement will go out shortly. The individual responsible for sending out the notice is on Pacific time and is not in the office yet. This upload to SDLC only just completed over the weekend. Regards PJones JD Neumann wrote:> I will check on this and see why it was not sent. > J.D. > > Wojciech Turek wrote: > >> Indeed you right! I used to get messages from lustre-announce list >> about such events but it seems it didn''t work this time. >> >> Thanks >> >> JD Neumann wrote: >> >>> !.6.6 is released and should be available from the down load center. >>> J.D. >>> >>> Wojciech Turek wrote: >>> >>>> I have some spare hardware (20TB of storage and several PE2950 >>>> servers) and I would like to use it as a test platform for new >>>> lustre version. I noticed that some people talk about testing beta >>>> version of 1.6.6. Can some one tell me where could I obtain rc >>>> version of 1.6.6 ? >>>> >>>> Many thanks, >>>> >>>> Mag Gam wrote: >>>> >>>>> Thankyou very much. I have done this for 48 hours now and no problem >>>>> to report. The new rc version is remarkable and the statahead problems >>>>> aren''t occuring anymore! >>>>> >>>>> Thanks again >>>>> >>>>> >>>>> On Sun, Oct 12, 2008 at 2:12 AM, Andreas Dilger <adilger at sun.com> >>>>> wrote: >>>>> >>>>> >>>>>> On Oct 11, 2008 12:15 -0400, Mag Gam wrote: >>>>>> >>>>>> >>>>>>> I meant to say, 5 clients running 1.6.6 and the remainder running >>>>>>> 1.6.5.1 >>>>>>> >>>>>>> >>>>>>> On Sat, Oct 11, 2008 at 9:14 AM, Mag Gam <magawake at gmail.com> wrote: >>>>>>> >>>>>>> >>>>>>>> Would there be a problem if we have 5 servers running 1.6.6 and the >>>>>>>> remainder of them running 1.6.5.1? Is there a chance of data >>>>>>>> corruption? >>>>>>>> >>>>>>>> >>>>>> No, we always test 1.x.y against 1.x.y-1 and 1.x-1.latest during our >>>>>> QA process. In general there should never be any chance of data >>>>>> corruption >>>>>> no matter what two releases you try to use together, because if there >>>>>> is (and we know about it, of course) then those versions would >>>>>> explicitly >>>>>> be detected as non-interoperable at connect time. >>>>>> >>>>>> Cheers, Andreas >>>>>> -- >>>>>> Andreas Dilger >>>>>> Sr. Staff Engineer, Lustre Group >>>>>> Sun Microsystems of Canada, Inc. >>>>>> >>>>>> >>>>>> >>>>>> >>>>> _______________________________________________ >>>>> Lustre-discuss mailing list >>>>> Lustre-discuss at lists.lustre.org >>>>> http://lists.lustre.org/mailman/listinfo/lustre-discuss >>>>> >>>>> >>>> ------------------------------------------------------------------------ >>>> >>>> >>>> _______________________________________________ >>>> Lustre-discuss mailing list >>>> Lustre-discuss at lists.lustre.org >>>> http://lists.lustre.org/mailman/listinfo/lustre-discuss >>>> >>>> > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss at lists.lustre.org > http://lists.lustre.org/mailman/listinfo/lustre-discuss > >