+++ This bug was initially created as a clone of Bug #979910 +++ Description of problem: with readdirplus support, kernel may not send lookup before a path based fop. Though readdirplus of each xlators do some things done in lookup, its not complete enough to replace a lookup. Hence its necessary that a lookup should be sent before we go ahead with path based fop. Version-Release number of selected component (if applicable): mainline How reproducible: consistently Steps to Reproduce: conifiguration: Two node replicate, fuse with readdirplus support (RHEL 6.4) 1. mount a glusterfs volume with existing data. 2. start copying from glusterfs mountpoint (it doesn't matter whether destination is glusterfs or not). 3. kill first (0th) brick of afr. Actual results: cp fails with ENOTCONN errors Expected results: cp shouldn't fail since 2nd brick is up and running. Additional info: --- Additional comment from Anand Avati on 2013-07-01 03:03:33 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#3) for review on master by Raghavendra G (rgowdapp) --- Additional comment from Anand Avati on 2013-07-02 01:03:39 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#4) for review on master by Raghavendra G (rgowdapp) --- Additional comment from Anand Avati on 2013-07-23 07:19:38 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#5) for review on master by Raghavendra G (rgowdapp) --- Additional comment from Anand Avati on 2013-07-23 21:59:48 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#6) for review on master by Raghavendra G (rgowdapp) --- Additional comment from Anand Avati on 2013-07-23 22:02:28 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#7) for review on master by Raghavendra G (rgowdapp) --- Additional comment from Anand Avati on 2013-07-23 22:06:23 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#8) for review on master by Raghavendra G (rgowdapp) --- Additional comment from Anand Avati on 2013-07-23 22:11:26 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#9) for review on master by Raghavendra G (rgowdapp) --- Additional comment from Anand Avati on 2013-07-23 22:18:11 EDT --- REVIEW: http://review.gluster.org/5267 (mount/fuse: "resolve" linked inodes which have not been looked up.) posted (#10) for review on master by Raghavendra G (rgowdapp)
This bug is actually important for supporting readdirplus support in glusterfs, which is needed for performance. https://code.engineering.redhat.com/gerrit/#/c/11021/
Verified on: glusterfs 3.4.0.19rhs built on Aug 14 2013 00:11:42 Tried to reproduce with the steps below: Steps to Reproduce: conifiguration: Two node replicate, fuse with readdirplus support (RHEL 6.4) 1. mount a glusterfs volume with existing data. 2. start copying from glusterfs mountpoint (it doesn't matter whether destination is glusterfs or not). 3. kill first (0th) brick of afr. Actual results: cp fails with ENOTCONN errors ====================== Unable to reproduce, marking verified.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. http://rhn.redhat.com/errata/RHBA-2013-1262.html