Pranith Kumar Karampuri

> > > > @xhernandez As per my recent understanding of crawling on a disk filesystem, this feature will improve performance of readdirp only when a directory has < 128KB...

@mohit84 I think readdir-ahead has bugs where it doesn't consider all xattrs that the parent xlators are querying for which leads to issues (especially with shard). We probably should fix...

@jirireischig The program you gave is performing writes where as the logs you mentioned are doing READs. Are you sure the log messages correspond to the program you mentioned in...

@jirireischig This amplifies my suspicion. The file is opened for O_WRONLY so reads on the file would fail with EBADFD. I wonder why reads are sent on this fd. Is...

@jirireischig Basically we need to know why reads are sent on a file opened for Writing. If you could send the bricks logs as well along with profile info output...

@jirireischig Looks like a bug in fuse-kernel? When I attach the process to gdb while testing it, fuse-kernel seems to be sending reads on a file opened with 'O_WRONLY' If...

Found a similar problem from 6 years ago! http://fuse.996288.n3.nabble.com/PATCH-fuse-fix-fuse-writeback-problem-with-O-WRONLY-td12699.html

@jirireischig You are correct. When I am debugging in gdb, if I put a breakpoint, it fails at the first write itself. I think kernel fuse is sending read when...

@jirireischig Could you follow up with kernel fuse mailing list as well? Let me include @amarts also in case he knows more about this.You can probably point to the link...

> > @jirireischig You are correct. When I am debugging in gdb, if I put a breakpoint, it fails at the first write itself. I think kernel fuse is sending...