5.6.4 HP StoreAll Storage Release Notes (AW549-96075, June 2013)

• When Cygwin was used to create a hard link on a CIFS/SMB share, the CIFS/SMB implementation
did not report that hard links are not supported.
• If a CIFS/SMB client had a file open on a StoreAll CIFS/SMB share and the client then lost network
access without closing the file gracefully, the file could not be opened by other CIFS/SMB clients
until a timeout period expired. This condition has been resolved by changing the TCP timeouts
used by the CIFS/SMB implementation. If you need to further adjust the timeout periods, locate
the following parameters in the Likewise registry:
"TcpKeepaliveCnt"=dword:00000009
"TcpKeepaliveIdle"=dword:00000618
"TcpKeepaliveIntvl"=dword:0000001e
The parameters are:
◦ TcpKeepaliveIdle: specifies how long the keepalive routines wait before sending the
first keepalive probe.
◦ TcpKeepaliveIntvl: When the keepalive time is reached, specifies how often to resend
a probe.
◦ TcpKeepaliveCnt: Specifies the number of probes to send with no ACK response received
before marking the connection as broken.
Try setting the values as follows, and adjust them further as needed:
"TcpKeepaliveCnt"=dword:00000005
"TcpKeepaliveIdle"=dword:00000010
"TcpKeepaliveIntvl"=dword:00000002
The Likewise registry is located at /opt/likewise/bin/lw-edit-reg. After adjusting the
values, restart the CIFS/SMB service.
Remote Replication
• When files were deleted on the source, remote replication took too long to delete them on the
destination.
• A CRR task remained in a persistent running state and had to be terminated manually.
Authentication
• Multiple unnecessary messages reporting Failed to authenticate user (name = '')
were logged in /var/log/messages.
• Incorrect invalid password errors caused the AD account to be locked.
• NTLMv2 authentication did not work correctly.
• When a user was removed from the builtin\administrators group, the remaining SID
could not be removed.
X9000 clients
• When files were modified with a StoreAll Windows client, the modifications were not replicated
to the target cluster.
• A race in an RPC request caused a StoreAll client to fail.
Other operations
• A server failed with the error ASSERT[atomic_read(&(*pfs)->ref.ref_counter) !=
0] failed: Too many unreferencing.
• The fmbackup.zip file could become corrupted.
• Old snapshots were not deleted, preventing the snapshot retention policy from working properly.
5.6.1 release 9