Hi Milosz, On Fri, 17 May 2013, Milosz Tanski wrote: > Hello there, > > I'm working on netfs-fscache support to the kernel ceph filesystem > client. The other filesystems in the kernel that have implemented it > are NFS and AFS. Awesome! > In the cache of a inode, how do I tell that's I'm the only person that > has this file open. So I can tell that I can safely cache the file. > After reading the code I imagine it has something to do with the > CEPH_CAP_FILE_SHARED, CEPH_CAP_FILE_EXCL and CEPH_CAP_FILE_CACHE. But > I wasn't able to find enough info to know that I've implemented my > cache invalidation correctly. FILE_CACHE means that we are allowed to cache what we read. The way to invalidate is to look at the rdcache_gen field in the ceph inode struct; if it has increased, that FILE_CACHE cap has been revoked and reissued, and what we had from before is stale. Does the cache persist across restarts/remounts? That is a bit more difficult, although it's possible we have a version field we can use for that as well... sage > To give you a little background. I'm doing this work because I think > it has great potential for our workload. We're using cephfs as a > backing store for our distributed database. Our frontend nodes (the > ones that perform the querying) contain decent size SSD drives. With > our primarily read workload and having a high rate of cachability for > file pages that contain indexes, data extents, ISAM trees. It would be > nice performance improvement. > > Thanks, > - Milosz > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
