Created attachment 585124 [details] Automation test run result Description of problem: ------------------------- files are created and are assigned gfid's on the bricks from the backed.Create a volume with the same bricks where the files were created.Create a nfs mount to the volume and look up on mount point should force merge these files onto the bricks. But the lookup is returning "No such file or directory" Error. Version-Release number of selected component (if applicable): 3.3.0qa41 How reproducible: often Steps to Reproduce: 1. create a file 'f1' on brick1 and 'f2' on brick2 from the backend 2. assign gfid's to the files 'f1' and 'f2' 3. create a replicate volume (1x2) with brick1 and brick2 4. create a nfs mount. 5. execute 'ls -l' from nfs mount Actual results: -------------- 2012-05-17 11:16:18,216 DEBUG : /home/tester/Testing/automation/libs/utils/hostutils.py execute_command 184 - '192.168.1.124: Executing Command: cd /mnt/nfsc1 ; ls -l' 2012-05-17 11:16:18,343 DEBUG : /home/tester/Testing/automation/libs/utils/atfutils.py print_stdout 154 - 'total 0 ' 2012-05-17 11:16:18,343 DEBUG : /home/tester/Testing/automation/libs/utils/atfutils.py print_stdout 154 - '-????????? ? ? ? ? ? f1 ' 2012-05-17 11:16:18,344 DEBUG : /home/tester/Testing/automation/libs/utils/atfutils.py print_stderr 160 - 'ls: cannot access f1: No such file or directory Expected results: ------------------- Should force merge of files f1 and f2. Additional Info:- -------------------- The same test works fine on fuse mount.
Created attachment 585126 [details] nfs server log
happens on distribute as well. Assigning to Rajesh. gluster --mode=script volume create d1 192.168.122.1:/gfs/d1_0 volume create: d1: success: please start the volume to access data gluster --mode=script volume start d1 volume start: d1: success gluster --mode=script volume set d1 client-log-level DEBUG volume set: success gluster --mode=script volume set d1 brick-log-level DEBUG volume set: success mkdir: cannot create directory `/mnt/d1': File exists mount -t glusterfs 192.168.122.1:/d1 /mnt/d1 root - ~ 14:07:48 :) ⚡ root - ~ 14:08:00 :) ⚡ mount -t nfs 192.168.122.1:/d1 /mnt/client -o vers=3 root - ~ 14:08:09 :) ⚡ touch /gfs/d1_0/a root - ~ 14:08:18 :) ⚡ ls -l /mnt/client ls: /mnt/client/a: No such file or directory total 0 -rw-r--r-- 0 root root 0 Dec 11 14:08 a root - ~ 14:08:22 :) ⚡ ls -l /mnt/client ls: /mnt/client/a: No such file or directory total 0 -rw-r--r-- 0 root root 0 Dec 11 14:08 a root - ~ 14:08:22 :) ⚡ ls -l /mnt/client ls: /mnt/client/a: No such file or directory total 0 -rw-r--r-- 0 root root 0 Dec 11 14:08 a
To force merge a file created in the backend directly for use in NFS, we need to have the exact hard/soft link in the backend .glusterfs directory, since NFS uses anonymous fds to operate on files.
I'll close this as NOTABUG because this is not a valid use-case in current versions. We expect that the .glusterfs directory on the bricks have hardlinks to the files. NFS uses these hardlinks to stat() a file, if these links are missing, things tend to got wrong. Please re-open if this still is an issue and you have a real-world use-case.