[Gluster-users] Corrupted File readable via FUSE?

[Amar Tumballi Suryanarayan]

Hi David,

I guess https://review.gluster.org/#/c/glusterfs/+/21996/ helps to fix the
issue. I will leave it to Raghavendra Bhat to reconfirm.

Regards,
Amar

On Fri, Feb 1, 2019 at 8:45 PM David Spisla <spisla80 at gmail.com> wrote:

> Hello Gluster Community,
> I have got a 4 Node Cluster with a Replica 4 Volume, so each node has a
> brick with a copy of a file. Now I tried out the bitrot functionality and
> corrupt the copy on the brick of node1. After this I scrub ondemand and the
> file is marked correctly as corrupted.
>
> No I try to read that file from FUSE on node1 (with corrupt copy):
> $ cat file1.txt
> cat: file1.txt: Transport endpoint is not connected
> FUSE log says:
>
> *[2019-02-01 15:02:19.191984] E [MSGID: 114031]
> [client-rpc-fops_v2.c:281:client4_0_open_cbk] 0-archive1-client-0: remote
> operation failed. Path: /data/file1.txt
> (b432c1d6-ece2-42f2-8749-b11e058c4be3) [Input/output error]*
> [2019-02-01 15:02:19.192269] W [dict.c:761:dict_ref]
> (-->/usr/lib64/glusterfs/5.3/xlator/performance/quick-read.so(+0x7329)
> [0x7fc642471329]
> -->/usr/lib64/glusterfs/5.3/xlator/performance/io-cache.so(+0xaaf5)
> [0x7fc642682af5] -->/usr/lib64/libglusterfs.so.0(dict_ref+0x58)
> [0x7fc64a78d218] ) 0-dict: dict is NULL [Invalid argument]
> [2019-02-01 15:02:19.192714] E [MSGID: 108009]
> [afr-open.c:220:afr_openfd_fix_open_cbk] 0-archive1-replicate-0: Failed to
> open /data/file1.txt on subvolume archive1-client-0 [Input/output error]
> *[2019-02-01 15:02:19.193009] W [fuse-bridge.c:2371:fuse_readv_cbk]
> 0-glusterfs-fuse: 147733: READ => -1
> gfid=b432c1d6-ece2-42f2-8749-b11e058c4be3 fd=0x7fc60408bbb8 (Transport
> endpoint is not connected)*
> [2019-02-01 15:02:19.193653] W [MSGID: 114028]
> [client-lk.c:347:delete_granted_locks_owner] 0-archive1-client-0: fdctx not
> valid [Invalid argument]
>
> And from FUSE on node2 (with heal copy):
> $ cat file1.txt
> file1
>
> It seems to be that node1 wants to get the file from its own brick, but
> the copy there is broken. Node2 gets the file from its own brick with a
> heal copy, so reading the file succeed.
> But I am wondering myself because sometimes reading the file from node1
> with the broken copy succeed
>
> What is the expected behaviour here? Is it possibly to read files with a
> corrupted copy from any client access?
>
> Regards
> David Spisla
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> https://lists.gluster.org/mailman/listinfo/gluster-users



-- 
Amar Tumballi (amarts)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20190204/1cc0c247/attachment.html>