[PVFS-users] [Fwd: problem writing to one of the disks in a PVFS fs (fwd)]

[Rob Ross]

Hi Helen,

The file existing with a length of zero indicates that the file metadata 
was created but no data was written.

Are there errors in the mgr's logfile?

It appears that for whatever reason your client is unable to get to that 
particular server.

When a file is opened/created, the mgr returns an IP address to the client 
that the client will then use to access that server.  One possibility is 
that the IP address for that server in the .iodtab file is incorrect.  
Another is that the mgr can get to that iod on that address but the client 
cannot (different network connectivity).  Another possibility is that the 
routing was somehow screwed up on the client.  Another options is that a 
firewall is preventing traffic to the server.

Do any of those sound reasonable?

Rob

On Tue, 14 Sep 2004, Phil Carns wrote:

> I am forwarding along this message from Helen Katz, who is currently 
> unable to post to the list; please preserve the cc when responding.
> 
> thanks,
> -Phil
> 
> ---------- Forwarded message ----------
> Date: Thu, 09 Sep 2004 02:22:04 -0400
> From: pvfs-users-owner at beowulf-underground.org
> To: helen.katz at weizmann.ac.il
> Subject: problem writing to one of the disks in a PVFS fs
> 
> 
> Hello,
> We are using pvfs-1.5.4 and  pvfs-kernel-1.5.4 under RH-7.3
> without problems and we have several stable PVFS filesystems,for
> several years .
> In 5'th of Sept. I had this problem, which I solved by rebooting
> the mgr machine:
> In one of the pvfs's we have problem writing to one of the disks.
> The disk /data08 from within eio08 is part of /pvfsa disk,
> and it's the only one to which I can't write.
> The /pvfsa disk is build from disks in eio05-eio12
>   pvfsa     Size(Kb)         Used         Free   Use%
>     0      115377640      5596960    103919768     6% eio05
>     1      115380192      6626576    102892580     7% eio06
>     2      115377640    104998728      4518000    96% eio07
>     3      115380192        65068    109454088     1% eio08
>     4      115380192      7469356    102049800     7% eio09
>     5      115380192      8845144    100674012     9% eio10
>     6      115380192      6037972    103481184     6% eio11
>     7      115345508     10353536     99132668    10% eio12
> When I use u2p command to write directly to a specific disk
> in that pvfs , I got this reply:
> ---------------------------------------------------------
> [physdsk2/fhkatzh1]~ [114] u2p -b 3 u2p.zsh /pvfsa/zgrp/fhkatzh1/u2p.zsh3
> add_accesses: connect: Connection refused
> pvfs_write: build_rw_jobs failed
> pvfs_write: short write
> 1 node(s); ssize = 65536; buffer = 3165; 9.896MBps (3165 bytes total)
> build_simple_jobs: connect: Connection refused
> pvfs_close: build_simple_jobs failed
> 
> The file exists with the length of zero.
> 
> When I use 'pvstat'  command I got the following:
> [physdsk2/fhkatzh1]~ [130] pvstat /pvfsa/zgrp/fhkatzh1/u2p.zsh3
> /pvfsa/zgrp/fhkatzh1/u2p.zsh3: base = 3, pcount = 1, ssize = 65536
> build_simple_jobs: connect: Connection refused
> pvfs_close: build_simple_jobs failed
> 
> while to other disks in the same pvfs it shows correctly:
> [physdsk2/fhkatzh1]~ [137] pvstat /pvfsa/zgrp/fhkatzh1/u2p.zsh2
> /pvfsa/zgrp/fhkatzh1/u2p.zsh2: base = 2, pcount = 1, ssize = 65536
> ----------------------------------------------------------
> When I use other '-b X' except 3, the writing is performed without
> problems.
> Can you ,please, suggest were to look to resolv this
> 'Connection refused' ?
> Thank you,
>    Helen
> helen.katz at weizmann.ac.il
> Particle Physics Dept.
> Weizmann Institute of Science
> Tel:  972-08-9342629
> 
> 
> _______________________________________________
> PVFS-users mailing list
> PVFS-users at www.beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs-users
> 
>