I would if I could but modifying the RHEVH instances seems like a bad idea.... Scott H. Login, LLC On 7/1/15 12:02 PM, Niels de Vos wrote:> On Wed, Jul 01, 2015 at 10:29:37AM -0700, Scott Harvanek wrote: >> so 3.6.0.53 isn't going to be opversion 30600 like stated here? >> http://www.gluster.org/community/documentation/index.php/OperatingVersions >> >> I'd assume the opver would remain the same even with their patches. >> >> I want the server cluster opversion and the client opversion to match so >> certain performance translators can be used, currently 3.6.3 on the server >> and 3.6.0.53 on the client blocks performance translators such as >> write-behind. > If you do not use the Red Hat Gluster Storage product, you will not > really get support on the glusterfs-client packages in RHEL either. The > combination of the two versions is just not (well) tested. You are > probably better of to use the community glusterfs client packages for > accessing the Gluster volumes too. > > Have you considered this approach? If you want the full support on the > gluster client side, you will need to use the Red Hat Gluster Storage > product on the storage servers. Any difficulties with deploying or > packaging the RHEL/RHGS product combination should get discussed with > your Red Hat support contacts. They should be able to explain how to do > it right, and/or get fixes in an updated version. > > Thanks, > Niels > >> Scott H. >> Login, LLC >> >> On 7/1/15 10:25 AM, Atin Mukherjee wrote: >>> -Atin >>> Sent from one plus one >>> On Jul 1, 2015 10:49 PM, "Scott Harvanek" <scott.harvanek at login.com >>> <mailto:scott.harvanek at login.com>> wrote: >>>> The client would be the lesser version in this case. Server @ 3.6.3, >>> Client @ 3.6.0 >>>> The only way to get them to match is to use redhat gluster storage where >>> you have RHEV hypervisors... which is a bit annoying. >>> Well if you are talking about rhs 3.6.0.53 as a client then definitely the >>> op version is not 30600 and also my question is why would you want to bump >>> down your server's cluster op version? >>>> Scott H. >>>> >>>> >>>> On 7/1/15 10:16 AM, Atin Mukherjee wrote: >>>>> -Atin >>>>> Sent from one plus one >>>>> On Jul 1, 2015 10:13 PM, "Scott Harvanek" <scott.harvanek at login.com >>> <mailto:scott.harvanek at login.com>> wrote: >>>>>> Scenario- >>>>>> >>>>>> Gluster bricks running Gluster 3.6.3, RHEL clients running RH >>> provided gluster 3.6.0.53 ( no way to change this, operating version 30600 >>> I presume ). >>>>> We do not encourage to use mix and match of upstream and downstream >>> versions. IIRC, op-version of 3.6.0.53 must be 30004 (please confirm) >>> which means client is higher version than server which is again not a >>> recommended setup. >>>>>> If we create a new cluster can the op-version be set down to 30600 on >>> the 3.6.3 bricks? - Currently - operating-version=30603 by default. >>>>>> -- >>>>>> Scott H. >>>>>> >>>>>> _______________________________________________ >>>>>> Gluster-users mailing list >>>>>> Gluster-users at gluster.org <mailto:Gluster-users at gluster.org> >>>>>> http://www.gluster.org/mailman/listinfo/gluster-users >>>> >> _______________________________________________ >> Gluster-users mailing list >> Gluster-users at gluster.org >> http://www.gluster.org/mailman/listinfo/gluster-users
Related question: What's the best ubuntu client version to use with RHS servers ? I have a mix of RH and Ubuntu clients, and the ubuntu clients are on the 3.5 ppa right now. I haven't seen any problems yet, but i would like to stick to best practices, if there are any. Once RHS 3.1 is rolled out, which I guess is based on 3.7 upstream, I was planning to upgrade clients to 3.6 ppa. i.e. I'm following the one major version older than the server heuristic. On Wed, Jul 1, 2015 at 12:09 PM, Scott Harvanek <scott.harvanek at login.com> wrote:> I would if I could but modifying the RHEVH instances seems like a bad > idea.... > > Scott H. > Login, LLC > > > On 7/1/15 12:02 PM, Niels de Vos wrote: > >> On Wed, Jul 01, 2015 at 10:29:37AM -0700, Scott Harvanek wrote: >> >>> so 3.6.0.53 isn't going to be opversion 30600 like stated here? >>> >>> http://www.gluster.org/community/documentation/index.php/OperatingVersions >>> >>> I'd assume the opver would remain the same even with their patches. >>> >>> I want the server cluster opversion and the client opversion to match so >>> certain performance translators can be used, currently 3.6.3 on the >>> server >>> and 3.6.0.53 on the client blocks performance translators such as >>> write-behind. >>> >> If you do not use the Red Hat Gluster Storage product, you will not >> really get support on the glusterfs-client packages in RHEL either. The >> combination of the two versions is just not (well) tested. You are >> probably better of to use the community glusterfs client packages for >> accessing the Gluster volumes too. >> >> Have you considered this approach? If you want the full support on the >> gluster client side, you will need to use the Red Hat Gluster Storage >> product on the storage servers. Any difficulties with deploying or >> packaging the RHEL/RHGS product combination should get discussed with >> your Red Hat support contacts. They should be able to explain how to do >> it right, and/or get fixes in an updated version. >> >> Thanks, >> Niels >> >> Scott H. >>> Login, LLC >>> >>> On 7/1/15 10:25 AM, Atin Mukherjee wrote: >>> >>>> -Atin >>>> Sent from one plus one >>>> On Jul 1, 2015 10:49 PM, "Scott Harvanek" <scott.harvanek at login.com >>>> <mailto:scott.harvanek at login.com>> wrote: >>>> >>>>> The client would be the lesser version in this case. Server @ 3.6.3, >>>>> >>>> Client @ 3.6.0 >>>> >>>>> The only way to get them to match is to use redhat gluster storage >>>>> where >>>>> >>>> you have RHEV hypervisors... which is a bit annoying. >>>> Well if you are talking about rhs 3.6.0.53 as a client then definitely >>>> the >>>> op version is not 30600 and also my question is why would you want to >>>> bump >>>> down your server's cluster op version? >>>> >>>>> Scott H. >>>>> >>>>> >>>>> On 7/1/15 10:16 AM, Atin Mukherjee wrote: >>>>> >>>>>> -Atin >>>>>> Sent from one plus one >>>>>> On Jul 1, 2015 10:13 PM, "Scott Harvanek" <scott.harvanek at login.com >>>>>> >>>>> <mailto:scott.harvanek at login.com>> wrote: >>>> >>>>> Scenario- >>>>>>> >>>>>>> Gluster bricks running Gluster 3.6.3, RHEL clients running RH >>>>>>> >>>>>> provided gluster 3.6.0.53 ( no way to change this, operating version >>>> 30600 >>>> I presume ). >>>> >>>>> We do not encourage to use mix and match of upstream and downstream >>>>>> >>>>> versions. IIRC, op-version of 3.6.0.53 must be 30004 (please confirm) >>>> which means client is higher version than server which is again not a >>>> recommended setup. >>>> >>>>> If we create a new cluster can the op-version be set down to 30600 on >>>>>>> >>>>>> the 3.6.3 bricks? - Currently - operating-version=30603 by default. >>>> >>>>> -- >>>>>>> Scott H. >>>>>>> >>>>>>> _______________________________________________ >>>>>>> Gluster-users mailing list >>>>>>> Gluster-users at gluster.org <mailto:Gluster-users at gluster.org> >>>>>>> http://www.gluster.org/mailman/listinfo/gluster-users >>>>>>> >>>>>> >>>>> _______________________________________________ >>> Gluster-users mailing list >>> Gluster-users at gluster.org >>> http://www.gluster.org/mailman/listinfo/gluster-users >>> >> > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20150701/5942d83c/attachment.html>
On Wed, Jul 01, 2015 at 12:09:23PM -0700, Scott Harvanek wrote:> I would if I could but modifying the RHEVH instances seems like a bad > idea....Ah, if it is on the RHEV-H and not a RHEL-H (or however that is called) you definitely are very restricted... Messing around with the op-version is not really nice, but that might be the easiest approach indeed. Atin surely is the one of the right guys to help out with that. Fortunately, the next Red Hat Gluster Storage packages will be based on glusterfs-3.7 with only minor (and compatible!) changes. This is supposed to prevent difficulties like this that are caused by the RHGS glusterfs-3.6 version. Niels> > Scott H. > Login, LLC > > On 7/1/15 12:02 PM, Niels de Vos wrote: > >On Wed, Jul 01, 2015 at 10:29:37AM -0700, Scott Harvanek wrote: > >>so 3.6.0.53 isn't going to be opversion 30600 like stated here? > >>http://www.gluster.org/community/documentation/index.php/OperatingVersions > >> > >>I'd assume the opver would remain the same even with their patches. > >> > >>I want the server cluster opversion and the client opversion to match so > >>certain performance translators can be used, currently 3.6.3 on the server > >>and 3.6.0.53 on the client blocks performance translators such as > >>write-behind. > >If you do not use the Red Hat Gluster Storage product, you will not > >really get support on the glusterfs-client packages in RHEL either. The > >combination of the two versions is just not (well) tested. You are > >probably better of to use the community glusterfs client packages for > >accessing the Gluster volumes too. > > > >Have you considered this approach? If you want the full support on the > >gluster client side, you will need to use the Red Hat Gluster Storage > >product on the storage servers. Any difficulties with deploying or > >packaging the RHEL/RHGS product combination should get discussed with > >your Red Hat support contacts. They should be able to explain how to do > >it right, and/or get fixes in an updated version. > > > >Thanks, > >Niels > > > >>Scott H. > >>Login, LLC > >> > >>On 7/1/15 10:25 AM, Atin Mukherjee wrote: > >>>-Atin > >>>Sent from one plus one > >>>On Jul 1, 2015 10:49 PM, "Scott Harvanek" <scott.harvanek at login.com > >>><mailto:scott.harvanek at login.com>> wrote: > >>>>The client would be the lesser version in this case. Server @ 3.6.3, > >>>Client @ 3.6.0 > >>>>The only way to get them to match is to use redhat gluster storage where > >>>you have RHEV hypervisors... which is a bit annoying. > >>>Well if you are talking about rhs 3.6.0.53 as a client then definitely the > >>>op version is not 30600 and also my question is why would you want to bump > >>>down your server's cluster op version? > >>>>Scott H. > >>>> > >>>> > >>>>On 7/1/15 10:16 AM, Atin Mukherjee wrote: > >>>>>-Atin > >>>>>Sent from one plus one > >>>>>On Jul 1, 2015 10:13 PM, "Scott Harvanek" <scott.harvanek at login.com > >>><mailto:scott.harvanek at login.com>> wrote: > >>>>>>Scenario- > >>>>>> > >>>>>>Gluster bricks running Gluster 3.6.3, RHEL clients running RH > >>>provided gluster 3.6.0.53 ( no way to change this, operating version 30600 > >>>I presume ). > >>>>>We do not encourage to use mix and match of upstream and downstream > >>>versions. IIRC, op-version of 3.6.0.53 must be 30004 (please confirm) > >>>which means client is higher version than server which is again not a > >>>recommended setup. > >>>>>>If we create a new cluster can the op-version be set down to 30600 on > >>>the 3.6.3 bricks? - Currently - operating-version=30603 by default. > >>>>>>-- > >>>>>>Scott H. > >>>>>> > >>>>>>_______________________________________________ > >>>>>>Gluster-users mailing list > >>>>>>Gluster-users at gluster.org <mailto:Gluster-users at gluster.org> > >>>>>>http://www.gluster.org/mailman/listinfo/gluster-users > >>>> > >>_______________________________________________ > >>Gluster-users mailing list > >>Gluster-users at gluster.org > >>http://www.gluster.org/mailman/listinfo/gluster-users >