CIFS VFS: No response to cmd

Question

Dear All, We get in trouble since a couple of weeks without finding any workaround. Therefore i'm asking to the community. Environment: VNX5300. full iscsi connectivity 10Gb. block version 05.31.000.5.709 - file 7.0.51-3. Issue: We have dedicated a share on our VNX with is accessed by a linux server (Red Hat Enterprise 5.5), via samba-client (samba-client.x86_64 - 3.0.33-3.39.el5_8). As soon as we ran a script which created and/or verify a list of folder and then copy several files, the cifs client hang and get those following message in /var/log/message:... Jul 30 10:30:30 test-us1-prd-db-a01 kernel:  CIFS VFS: No response to cmd 47 mid 283 Jul 30 10:30:30 test-us1-prd-db-a01 kernel:  CIFS VFS: Write2 ret -11, wrote 0 Jul 30 10:30:36 test-us1-prd-db-a01 kernel:  CIFS VFS: Write2 ret -11, wrote 0 Jul 30 10:31:21 test-us1-prd-db-a01 kernel:  CIFS VFS: No response to cmd 47 mid 288 Jul 30 10:31:21 test-us1-prd-db-a01 kernel:  CIFS VFS: Write2 ret -11, wrote 0 Jul 30 10:31:25 test-us1-prd-db-a01 kernel:  CIFS VFS: Write2 ret -11, wrote 0 Jul 30 10:32:10 test-us1-prd-db-a01 kernel:  CIFS VFS: No response to cmd 47 mid 293 Jul 30 10:32:59 test-us1-prd-db-a01 kernel:  CIFS VFS: No response to cmd 47 mid 298 Jul 30 10:33:48 test-us1-prd-db-a01 kernel:  CIFS VFS: No response to cmd 47 mid 303 ... Consequences: No possibility to do anyelse on the share. But we are able to create manually some files and/or folder in it one by one. What we tried: We had tried to reproduce this issue on a centos, with disabling mtu (was set as 9000), reset every permission and put everyone/full control, recreated a file system from scratch, mount the share with very basic cifs option into fstab, change on the fly the cifs option into /proc/fs/cifs... and unfortunaltely each time we can reproduce the same behaviour. Conclusion: it seems that it's on high load where the samba client hang, but now I'm unable to look at something efficiently Any one have got already this kind of message; or get a way of troubleshooting. Thank for any feedback or ideas. any help would be appreciated ;-) Baud

Rainer_EMC · Answer

How about using NFS ? Anything in th data mover logs at the time the problem happens. Otherwise I suggest to open a service request and take a network trace.

SAMEERK1 · Answer

Hi, Check this link, let me know if this helps you. http://blog.dhampir.no/content/cifs-vfs-no-response-for-cmd-n-mid Sameer Kulkarni

Rainer_EMC · Answer

Sounds like a bug in the Linux CIFS client – I would suggest trying to update that

runteam · Answer

Thank you for your following.

We didn't plan to use NFS as the share will be also used by a windows client.

We also did a network trace without success, as we saw that the NAS didn't answer during the script execution.

All info found via internet to change the cifs option on the fly, as Sameer suggest, didn't work.

And finaly, the linux server is up to date.

We play around once again today and it seems finaly that the MTU disabling was not correctly taken into account, as we have to explicitly set into the network parameters "MTU=1500" to disable the jumbo frame.

Now we've got something which works. So we have to verify all network path and switches to see if you have not forget to configure jumbo frame somewhere.

Thank for support.
Baud

Rainer_EMC · Answer

Hmmh, If Jumbo Frames were the problem you should have seen it in the network stats

runteam · Answer

you're right Rainer,but some time, we shouldn't trust too much our network team.

VNX

Was this post helpful?