Hello, I am trying to couple OCFS2 with a storage virtualization environment to use features like mirroring and snapshots. Unfortunately it seems to be impossible for ocfs2console (and for mount.ocfs2) to distinguish between the original volume and its snapshot and ocffs2 stops the mount-process with the following messages: Jul 20 17:23:54 sinalco kernel: (5028,1):ocfs2_initialize_super:1395 max_slots for this device: 4 Jul 20 17:23:54 sinalco kernel: (5028,0):ocfs2_fill_super:642 ERROR: Unable to create per-mount debugfs root. Is it possible too change the uuid and other relevant parameters? Best Regards Andr? -- ------------------------------------- Dr.-Ing. Andr? Brinkmann Heinz Nixdorf Institute University of Paderborn F?rstenallee 11 33102 Paderborn Germany ------------------------------------- Email: brinkman at hni.upb.de Phone: +49-5251-60 63 42 Fax: +49-5251-60 63 51 -------------------------------------
Are you sure, that your snapshort/mirroring software can work on the cluster (shared disks)? I do not know such tools, except parts of Poliserver file system and (possible, with serious limitation) Veritas Volume Manager. OCFSv2 is new and not matured by itself; don't add extra instability into it. ----- Original Message ----- From: "Andre Brinkmann" <brinkman at hni.upb.de> To: <ocfs2-users at oss.oracle.com> Sent: Thursday, July 20, 2006 8:39 AM Subject: [Ocfs2-users] OCFS2 and Snapshots Hello, I am trying to couple OCFS2 with a storage virtualization environment to use features like mirroring and snapshots. Unfortunately it seems to be impossible for ocfs2console (and for mount.ocfs2) to distinguish between the original volume and its snapshot and ocffs2 stops the mount-process with the following messages: Jul 20 17:23:54 sinalco kernel: (5028,1):ocfs2_initialize_super:1395 max_slots for this device: 4 Jul 20 17:23:54 sinalco kernel: (5028,0):ocfs2_fill_super:642 ERROR: Unable to create per-mount debugfs root. Is it possible too change the uuid and other relevant parameters? Best Regards Andr? -- ------------------------------------- Dr.-Ing. Andr? Brinkmann Heinz Nixdorf Institute University of Paderborn F?rstenallee 11 33102 Paderborn Germany ------------------------------------- Email: brinkman at hni.upb.de Phone: +49-5251-60 63 42 Fax: +49-5251-60 63 51 ------------------------------------- _______________________________________________ Ocfs2-users mailing list Ocfs2-users at oss.oracle.com http://oss.oracle.com/mailman/listinfo/ocfs2-users
OCFS2 relies on the uniqueness of the uuid for it to distinguish between different volumes. One cannot mount two volumes having the same uuid on the same node. Infact, one should not do that across the cluster too, i.e., mount two different physical volumes having the same identical uuid. If you have to mirror and mount, mount it on another node in a different cluster. It could be a 1 node cluster too. Andre Brinkmann wrote:> Hello, > > I am trying to couple OCFS2 with a storage virtualization environment to > use features like mirroring and snapshots. Unfortunately it seems to be > impossible for ocfs2console (and for mount.ocfs2) to distinguish between > the original volume and its snapshot and ocffs2 stops the mount-process > with the following messages: > > Jul 20 17:23:54 sinalco kernel: (5028,1):ocfs2_initialize_super:1395 > max_slots for this device: 4 > Jul 20 17:23:54 sinalco kernel: (5028,0):ocfs2_fill_super:642 ERROR: > Unable to create per-mount debugfs root. > > Is it possible too change the uuid and other relevant parameters? > > Best Regards > > Andr? > >
Andre Brinkmann <brinkman at hni.upb.de> wrote:> I am trying to couple OCFS2 with a storage virtualization > environment to use features like mirroring and snapshots. > Unfortunately it seems to be impossible for ocfs2console (and > for mount.ocfs2) to distinguish between the original volume > and its snapshot and ocffs2 stops the mount-process with the > following messages:[snip]> Is it possible too change the uuid and other relevant > parameters?On Thu, Jul 20, 2006 at 09:29:29AM -0700, Alexei_Roudnev replied:> Are you sure, that your snapshort/mirroring software can work > on the cluster (shared disks)? I do not know such tools, > except parts of Poliserver file system and (possible, with > serious limitation) Veritas Volume Manager.I am sure he means that he is using something like an HP EVA or IBM DS4000 etc. that can do a snapshot of a LUN. This looks like a block for block copy of the original LUN, so if you try to mount it, the UUIDs etc. will be identical and OCFS2 will complain. i.e. the snapshot "software" is built into the Storage array and does not need to be aware of any clustering. It just copies (or copies on write) the data to another area of disk which can be made available to the original machine(s) or to other machines entirely. -- Michael Wood <mwood at icts.uct.ac.za>
Michael Wood wrote:> Andre Brinkmann <brinkman at hni.upb.de> wrote: > >> I am trying to couple OCFS2 with a storage virtualization >> environment to use features like mirroring and snapshots. >> Unfortunately it seems to be impossible for ocfs2console (and >> for mount.ocfs2) to distinguish between the original volume >> and its snapshot and ocffs2 stops the mount-process with the >> following messages: >> > [snip] > >> Is it possible too change the uuid and other relevant >> parameters? >> > > On Thu, Jul 20, 2006 at 09:29:29AM -0700, Alexei_Roudnev replied: > >> Are you sure, that your snapshort/mirroring software can work >> on the cluster (shared disks)? I do not know such tools, >> except parts of Poliserver file system and (possible, with >> serious limitation) Veritas Volume Manager. >> > > I am sure he means that he is using something like an HP EVA or > IBM DS4000 etc. that can do a snapshot of a LUN. This looks > like a block for block copy of the original LUN, so if you try > to mount it, the UUIDs etc. will be identical and OCFS2 will > complain. > > i.e. the snapshot "software" is built into the Storage array and > does not need to be aware of any clustering. It just copies (or > copies on write) the data to another area of disk which can be > made available to the original machine(s) or to other machines > entirely. > >In my case, I am using the built-in disks of my cluster, export them via iscsi-target software and couple them via open-iscsi and our own storage virtualization and snapshot software. For OCFS2 (running on the same nodes), there is no distinction between our storage implementation and an HP EVA or IBM DS4000. Therefore your description hits the critical point. Whatever external snapshot solution is used with OCFS2, you need to be able to change the uuid to use the snapshot and the original volume in the same cluster. Based on the remarks of Joel, I am currently looking inside tunefs.ocfs and think that I will have the feature implemented this evening for evaluation purposes. -- ------------------------------------- Dr.-Ing. Andr? Brinkmann Heinz Nixdorf Institute University of Paderborn F?rstenallee 11 33102 Paderborn Germany ------------------------------------- Email: brinkman at hni.upb.de Phone: +49-5251-60 63 42 Fax: +49-5251-60 63 51 -------------------------------------