Powerconnect error on a stack

hi,

We have a problem with one on our new powerconnect 6248.

Our configuration :

- 2 6248 on a stack (default configuration, version 3.3.6.4 both, no routing, one lag directly connected to another stack who works fine)

- Everything worked fine (during 1 week)

Problem :

- The unit 1 (a stack member) leave the stack and passed at the state unavailable, we don't inderstand why.

Some ouput :

- System > General > Versions:

1 Not Available 0.0.0.0 image1 image1

2 3.3.6.4 3.3.6.4 3.3.3.3 image1 image1 

-  System > Stacking > Stack Summary :

1 Unassigned   PCT6248   Not present 0.0.0.0 
2 Management Unit   PCT6248 PCT6248 OK 3.3.6.4 

- logs:

Error  DNS_CLIEN DNS Client: osapiSocketRecvFrom returned error for addr. Indicates there is a stack error in receiving the DNS response packet from the server.

 We have some error like that, we finally disabled the dns entries (we don't need it anymore)

and  suddenly : 

9338  Notice  TRAPMGR Unit 1 is removed from the stack 
9339  Notice TRAPMGR Entity Database: Configuration Changed 
9340  Error  DRIVER ATP: TX timeout, seq 21910. cli 778. to 1(mac d0:67:e5:6f:6c:9) tx cnt 21. 

... some spanning tree entries

If someone know something about this error (firm current error?)...

Thank you for your help

Jeremy M.