CIFS - DNS Question

[Rahim Johnson]

We had an issue with Windows users losing the ability to connect to our CIFS share (HP-UX 11.2 in a SG Cluster, the CIFS package has its own IP for failover)

The users are authenticated via PAM/KRB5.

The problem was resolved with a reboot of our primary Win2k3 DC. But we still do not know what caused the problem.

Here are some other factors:

1. The CIFS server on a different HP-UX server had no problems at all.

2. The Shares could be accessed by name and not by IP address, during the problem.

3. There was no other resolution problems on the network.



Here is the samba log error when the problem surfaced:

[2010/11/02 14:28:43, 0] rpc_client/cli_pipe.c:cli_rpc_pipe_open_noauth(2239)
cli_rpc_pipe_open_noauth: rpc_pipe_bind for pipe \NETLOGON failed with error NT_STATUS_BUFFER_TOO_SMALL
[2010/11/02 14:28:43, 0] rpc_client/cli_pipe.c:cli_rpc_pipe_open_schannel(2643)
cli_rpc_pipe_open_schannel: failed to get schannel session key from server XXX-DC1 for domain XXX.
[2010/11/02 14:28:43, 0] auth/auth_domain.c:connect_to_domain_password_server(111)
connect_to_domain_password_server: unable to open the domain client session to machine XXX-DC1. Error was : NT_STATUS_BUFFER
_TOO_SMALL.

So this appears to be a problem with this Unix Host only, but when troubleshooting these errors, even google stuttered, so I am just trying to see if anyone else has run into this.

Thanks

Again, a Domain Controller reboot fixed it, but I am trying to isolate the cause.

[Rahim Johnson]

Additional Note:

There have been no major system changes and the configuration has not changed since implemented two years ago.

[Hakki Aydin Ucar]

I found that from google (may be you missed this one):

> NT_STATUS_BUFFER_TOO_SMALL

That's not really an error - that's the client trying to size the security descriptor buffer to use for the next query. Perfectly normal.

I am not an expert on CIFS but, may be you can go on with CIFS share configuration to isolate root cause on HP-UX server ?

[Bill Hassell]

You will see a long list of problems in these forums with Windows DNS servers, some due to setup, some due incomplete configurations and some due to reliability. If this is a production server, don't use DNS as your primary host/IP reference. Instead, change to use /etc/hosts by editing /etc/nsswtch.conf and put the relevant IP and hostnames in that file. The 2 entries should be:

hosts: files [NOTFOUND=continue UNAVAIL=continue] dns
ipnodes: files [NOTFOUND=return UNAVAIL=return TRYAGAIN=return]

[Rahim Johnson]

I will take a look at that. Thanks