Let me first say that we are very happy with the high quality and the performance of aufs and we are very glad for all the work you put into maintaining and improving it!
At Cern we have two different contexts in which we use aufs. 1) Aufs creates a writable root file system for a virtual machine that is used as a cloud image and on users' laptops to crunch physics data. The read-only branch is a fuse-based network file system, the read-write branch is on storage that is local to the virtual machine [1]. For the virtual machine, we use a vanilla kernel and I guess that we can switch relatively quickly from 3.10 to 3.14. 2) Aufs creates the writable interface to an otherwise read-only file system, which is used by the LHC experiments, in general in high-energy physics and in other fields in order to distribute scientific software to data centers around the world. All normal computers mount the read-only file system. A few special computers are used to stage updated and new software versions. On these special computers, we let AUFS write all modifications into a temporary directory, which we then scan to pickup the modifications. The new and modified files are merged with the existing storage of our file system [2]. The special machines are operated in several, independent data centers. Most of them have RHEL on their machines (or CentOS or Scientific Linux). Since Redhat does not provide aufs, we rebuild the Redhat kernel with aufs patches so that it is easy for administrators to operate the special machines with their normal configuration management tools (puppet, chef, or whatever they like to use) [3]. The aufs version that matches closest the RHEL6 kernel is aufs2.1-32. For the new RHEL7 kernel I made some initial tests with the aufs3-3.10 patch set and that worked just fine. I'm afraid that we at Cern have to support aufs on RHEL7 for the next few years. I think there would not be a need for backported features. The current aufs feature set is plenty! But in case a critical security fix comes up or in case the vanilla kernel is changed in a way that the aufs patch set doesn't compile anymore, any help is much appreciated. For instance, we are very grateful that you introduced the aufs3.10.x patch set. Now, this our "purely egoistic" point of view. I fully understand that much work is involved in maintaining many different versions. Please consider this lengthy email merely as an explanation of our current use of aufs. I was also thinking, if there are automatic build procedures which we could follow, we can certainly spare machines in our automatic build and test system infrastructure for aufs builds. Best regards, Jakob [1] http://iopscience.iop.org/1742-6596/513/3/032009 [2] http://iopscience.iop.org/1742-6596/396/5/052013 [3] I think I announced the aufs patched RHEL6 rpms sometime earlier on this mailing list. Anyway, here's the link again in case it is useful for someone else: http://cvmrepo.web.cern.ch/cvmrepo/yum/cvmfs-kernel/EL/6/x86_64/ ------------------------------------------------------------------------------ Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds
