[Gluster-users] favorite-child and self-heal BUG ?

[Keith Freedman]

>As you can see the files replicated themselves but only after the 
>third attempt to open the file on a client. As I understand I 
>shouldn't remove the file directly on the server but such situation 
>might occur. Perhaps the scenarion should be addressed somehow?

yes, I'd agree with this.. the I/O error shouldn't happen. gluster 
should figure out the files missing, determine whether or not it 
should be there or not there, and then fix before reporting back an I/O error.

again, what version of gluster are you using.  I remember running 
into this a while ago, and the problem was resolved.

what I'm seeing in 1.4.0.patch-824 is the following:

I removed a file from the underlying filesystem on server A.
I did an LS of the mounted filesystem on server a, and it was there, 
I did a cat of the file and it was there.
but when I look for it in the underlying filesystem it's not there.
so, it knows it's not on the local volume, and it gets it from the 
other server, but it doesn't auto-heal the file back?
this seems like a bug, but I defintiely didn't get the I/O error you saw.





>logs:
>2009-01-09 09:20:19 E [afr-self-heal-data.c:778:afr_sh_data_fix] 
>afr: Unable to resolve conflicting data of /8. Please resolve 
>manually by deleting the file /8 from all but the preferred 
>subvolume. Please consider 'option favorite-child <>'
>2009-01-09 09:20:19 W [afr.c:618:afr_open] afr: returning EIO, file 
>has to be manually corrected in backend
>2009-01-09 09:20:19 E [fuse-bridge.c:662:fuse_fd_cbk] 
>glusterfs-fuse: 1091: OPEN() /8 => -1 (Input/output error)
>2009-01-09 09:20:20 W [afr.c:618:afr_open] afr: returning EIO, file 
>has to be manually corrected in backend
>2009-01-09 09:20:20 E [fuse-bridge.c:662:fuse_fd_cbk] 
>glusterfs-fuse: 1092: OPEN() /8 => -1 (Input/output error)
>2009-01-09 09:20:21 W [afr-self-heal-data.c:764:afr_sh_data_fix] 
>afr: Picking favorite child client1 as authentic source to resolve 
>conflicting data of /8
>2009-01-09 09:20:21 W 
>[afr-self-heal-data.c:615:afr_sh_data_open_cbk] afr: sourcing file 
>/8 from client1 to other sinks
>2009-01-09 09:22:10 W 
>[afr-self-heal-common.c:843:sh_missing_entries_lookup_cbk] afr: path 
>/9 on subvolume client1 => -1 (No such file or directory)
>2009-01-09 09:22:10 E [afr-self-heal-data.c:778:afr_sh_data_fix] 
>afr: Unable to resolve conflicting data of /9. Please resolve 
>manually by deleting the file /9 from all but the preferred 
>subvolume. Please consider 'option favorite-child <>'
>2009-01-09 09:22:10 W [afr.c:618:afr_open] afr: returning EIO, file 
>has to be manually corrected in backend
>2009-01-09 09:22:10 E [fuse-bridge.c:662:fuse_fd_cbk] 
>glusterfs-fuse: 1114: OPEN() /9 => -1 (Input/output error)
>2009-01-09 09:22:11 W [afr.c:618:afr_open] afr: returning EIO, file 
>has to be manually corrected in backend
>2009-01-09 09:22:11 E [fuse-bridge.c:662:fuse_fd_cbk] 
>glusterfs-fuse: 1115: OPEN() /9 => -1 (Input/output error)
>2009-01-09 09:22:12 W [afr-self-heal-data.c:764:afr_sh_data_fix] 
>afr: Picking favorite child client1 as authentic source to resolve 
>conflicting data of /9
>2009-01-09 09:22:12 W 
>[afr-self-heal-data.c:615:afr_sh_data_open_cbk] afr: sourcing file 
>/9 from client1 to other sinks
>
>
>
>
>
>
>_______________________________________________
>Gluster-users mailing list
>Gluster-users at gluster.org
>http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users