[Gluster-devel] Rehashing because st_nlink less than dentry maps

[Yann Dupont]

Raghavendra G a écrit :
> Hi Yann,

Hello. Sorry for the late answer.
> Is it possible to attach the complete log file from client side? Also 
> can you please provide the following information?

first the context : I'm trying to put a server for freenx (no machine). 
There is lots of accounts. Home directory are created on the fly with pam.

Glusterfs is hosted on 3 nodes, Each has 2 volumes.
each volume is mirrored with afr.

The 3 subvolumes are unified.

The whole is used on a 4th machine.



I've turne all optimisations off for the moment.
Here is the glusterfs-client.vol

# file: /etc/glusterfs/glusterfs-client.vol

## 1er groupe.

volume remote1a
  type protocol/client
  option transport-type tcp/client
  option remote-host gluster1.univ-nantes.prive
#  option window-size 2097152
  option remote-subvolume brick3
end-volume

volume remote1b
  type protocol/client
  option transport-type tcp/client
  option remote-host gluster2.univ-nantes.prive
#  option window-size 2097152
  option remote-subvolume brick3
end-volume

volume mirror1
  type cluster/afr
  subvolumes remote1a remote1b
end-volume

#### 2eme groupe.

volume remote2a
  type protocol/client
  option transport-type tcp/client
  option remote-host gluster1.univ-nantes.prive
#  option window-size 2097152
  option remote-subvolume brick4
end-volume

volume remote2b
  type protocol/client
  option transport-type tcp/client
  option remote-host gluster3.univ-nantes.prive
#  option window-size 2097152
  option remote-subvolume brick3
end-volume

volume mirror2
  type cluster/afr
  subvolumes remote2a remote2b
end-volume
volume remote3a
  type protocol/client
  option transport-type tcp/client
  option remote-host gluster3.univ-nantes.prive
#  option window-size 2097152
  option remote-subvolume brick4
end-volume

volume remote3b
  type protocol/client
  option transport-type tcp/client
  option remote-host gluster2.univ-nantes.prive
#  option window-size 2097152
  option remote-subvolume brick4
end-volume

volume mirror3
  type cluster/afr
  subvolumes remote3a remote3b
end-volume

### Assemblage avec unify.


volume remote-ns
  type protocol/client
  option transport-type tcp/client
  option remote-host gluster1.univ-nantes.prive
  option remote-subvolume ns2
end-volume

volume unify0
  type cluster/unify
  option scheduler rr # round robin
  option namespace remote-ns
  subvolumes mirror1 mirror2 mirror3
end-volume






the volume is mounted OK. df, ls, etc is working OK.


The I tru to access the server with nxclient. This is working via ssh.
The login fail when home is on glusterfs. It's fine when the home 
directory is on local disk.

Here are the logs. 2 try.  I've turned debug on.
I've attached the file, this is long.








> * Filesystem operation which caused this hang.
Only access to the server via nxclient.
nxclient is connecting via ssh and forks a process, as far as I can see.

> * Is any other process accessing the same file (or any other operation 
> being performed on the same file), for which fuse_fd_cbk said that 
> open got EINTR?
Not that I know.

>
> In general any information describing about actions being performed on 
> the glusterfs during hang would be helpful.
>
The server is idle & I'm not performing anything else.

Regards,

-- 
Yann Dupont - Service IRTS, DSI Université de Nantes
Tel : 02.51.12.53.91 - Mail/Jabber : Yann.Dupont at univ-nantes.fr

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: glusterfs.log
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20080526/17a6b6c0/attachment-0003.ksh>