Re: [Gluster-users] GlusterFS over S3FS
And, like I just mentioned on the IRC channel, if someone had the skills a storage/s3 translator might be interesting (would probably sit in place of storage/posix). Could even likely copy/paste most of it from S3QL or S3FS. On 06/15/2016 04:32 AM, Niklaas Baudet von Gersdorff wrote: Vincent Miszczak [2016-06-15 10:27 +] : I would like to combine Glusterfs with S3FS. [...] I also have the idea to test this with Swift object storage. Advises are welcome. Never tried this before. Maybe S3QL [1] works since it "is a standard conforming, full featured UNIX file system that is conceptually indistinguishable from any local file system". 1: https://bitbucket.org/nikratio/s3ql/ The entire approach sounds a bit hackish to me though. :-) Niklaas ___ Gluster-users mailing list Gluster-users@gluster.org http://www.gluster.org/mailman/listinfo/gluster-users ___ Gluster-users mailing list Gluster-users@gluster.org http://www.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] GlusterFS over S3FS
Glusterfs requires extended attributes. Does S3FS support xattrs? On Thu, Jun 16, 2016 at 1:14 AM, Vincent Miszczak wrote: > Hello, > > > Thank you for pointing this project, gonna try this. > > > The idea behind what I described, is to provide SMB shares with automatic > placement based on usage pattern. > > I work with large volumes, and only a fraction should have good (costly) > performance. The rest can be archived. > > > Archiving "the old way" (I mean manually moving the files to a cold > tier) is not convenient, it breaks URLs, unless someone has some tips about > this. > > > I'm able to do the scenario described with normal Gluster nodes, some > local with costly storage, some remotes with cheap storage. It's just > tiering. > > But I still have to manage the Linux behind the cold tier. Not interesting > to me as AWS, Google or whatever can provide cheap object storage without > maintenance (that means, less people in my organization to achieve the same > job). > > > Vincent > -- > *From:* gluster-users-boun...@gluster.org < > gluster-users-boun...@gluster.org> on behalf of Niklaas Baudet von > Gersdorff > *Sent:* Wednesday, June 15, 2016 1:32:58 PM > *To:* gluster-users@gluster.org > *Subject:* Re: [Gluster-users] GlusterFS over S3FS > > Vincent Miszczak [2016-06-15 10:27 +] : > > > I would like to combine Glusterfs with S3FS. > [...] > > I also have the idea to test this with Swift object storage. Advises are > welcome. > > Never tried this before. Maybe S3QL [1] works since it "is > a standard conforming, full featured UNIX file system that is > conceptually indistinguishable from any local file system". > > 1: https://bitbucket.org/nikratio/s3ql/ > > The entire approach sounds a bit hackish to me though. :-) > > Niklaas > > ___ > Gluster-users mailing list > Gluster-users@gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users > ___ Gluster-users mailing list Gluster-users@gluster.org http://www.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] GlusterFS over S3FS
Hello, Thank you for pointing this project, gonna try this. The idea behind what I described, is to provide SMB shares with automatic placement based on usage pattern. I work with large volumes, and only a fraction should have good (costly) performance. The rest can be archived. Archiving "the old way" (I mean manually moving the files to a cold tier) is not convenient, it breaks URLs, unless someone has some tips about this. I'm able to do the scenario described with normal Gluster nodes, some local with costly storage, some remotes with cheap storage. It's just tiering. But I still have to manage the Linux behind the cold tier. Not interesting to me as AWS, Google or whatever can provide cheap object storage without maintenance (that means, less people in my organization to achieve the same job). Vincent From: gluster-users-boun...@gluster.org on behalf of Niklaas Baudet von Gersdorff Sent: Wednesday, June 15, 2016 1:32:58 PM To: gluster-users@gluster.org Subject: Re: [Gluster-users] GlusterFS over S3FS Vincent Miszczak [2016-06-15 10:27 +] : > I would like to combine Glusterfs with S3FS. [...] > I also have the idea to test this with Swift object storage. Advises are > welcome. Never tried this before. Maybe S3QL [1] works since it "is a standard conforming, full featured UNIX file system that is conceptually indistinguishable from any local file system". 1: https://bitbucket.org/nikratio/s3ql/ The entire approach sounds a bit hackish to me though. :-) Niklaas ___ Gluster-users mailing list Gluster-users@gluster.org http://www.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] GlusterFS over S3FS
Vincent Miszczak [2016-06-15 10:27 +] : > I would like to combine Glusterfs with S3FS. [...] > I also have the idea to test this with Swift object storage. Advises are > welcome. Never tried this before. Maybe S3QL [1] works since it "is a standard conforming, full featured UNIX file system that is conceptually indistinguishable from any local file system". 1: https://bitbucket.org/nikratio/s3ql/ The entire approach sounds a bit hackish to me though. :-) Niklaas signature.asc Description: PGP signature ___ Gluster-users mailing list Gluster-users@gluster.org http://www.gluster.org/mailman/listinfo/gluster-users
[Gluster-users] GlusterFS over S3FS
Hello, I would like to combine Glusterfs with S3FS. The idea is to create a tiered volume composed of local network storage and remote object storage (something similar to vendor tiering solutions, but with some tiers being remote). I have successfully setup local Glusterfs part on one side, and S3FS mount on another side: -local volume is OK (create, start) -creating a Gluster volume on S3FS is OK -starting the volume on S3FS fails (/gluster/s3fs is symlinked to /s3fs) root@dfs:/gluster/s3fs# gluster volume create dfs dfs:/s3fs/test volume create: dfs: success: please start the volume to access data root@dfs:/gluster/s3fs# gluster volume start dfs volume start: dfs: failed: Commit failed on localhost. Please check log file for details. >From the log: [2016-06-15 09:36:13.289768] E [MSGID: 113082] [posix.c:6638:init] 0-dfs-posix: /s3fs/test: failed to set gfid [Le fichier existe] [2016-06-15 09:36:13.289831] E [MSGID: 101019] [xlator.c:433:xlator_init] 0-dfs-posix: Initialization of volume 'dfs-posix' failed, review your volfile again [2016-06-15 09:36:13.289840] E [graph.c:322:glusterfs_graph_init] 0-dfs-posix: initializing translator failed [2016-06-15 09:36:13.289847] E [graph.c:661:glusterfs_graph_activate] 0-graph: init failed It looks like S3FS is not compliant with Gluster expectation. Questions : Has someone tried to do similar things ? Is there any chance this kind of scenario can work ? I also have the idea to test this with Swift object storage. Advises are welcome. Vincent ___ Gluster-users mailing list Gluster-users@gluster.org http://www.gluster.org/mailman/listinfo/gluster-users