Highly Available CIFS configuration problem

[Oleg Podoukov]

Hi,
I need to run CIFS 2.1.1 with ServiceGuard in active-active mode in two nodes cluster. I did all the configurations exactly the same as Admin's guide says. I've changed path from /var/opt/samba/locks default directory to /var/opt/samba/first_samba_node/locks directory in smb.conf.first_samba_node file ("lock directory =" parameter). But smbd and nmbd daemons (smbd -D -l /var/opt/samba/first_samba_node/logs -s /etc/opt/samba/smb.conf.first_samba_node) continue creating smbd.pid and nmbd.pid files to /var/opt/samba/locks directory. So, samba.cntl and samba.mon scripts doesn't work properly. Does anybody know the solution?

Thank you in advance,
Oleg

[Steven E. Protter]

Servicegaurd has a log file. Please look at them and post the error text for help. Any time a monitor script fails the message is logged in a file in the /etc/cmcluster folder.

There may be useful information in the samba log, which is defined in smb.conf

SEP

[RolandH]

Have you checked that you have changed your parameter in the smb.conf file.

lock dir = /var/opt/samba/first_samba_node/locks

pid directory = /var/opt/samba/first_samba_node/pids

How are these parameters set in your control file, too?
SMBD_PID_FILE=
NMBD_PID_FILE=


Roland

[Oleg Podoukov]

Thanks a lot, Roland! Addition of "pid directory =" to smb.conf.node_name files helped. Admin's guide tells nothing about this.
But there is another bug. After cluster started, there are one nmbd and two smbd daemons running:
root 10112 1 0 14:44:19 ? 0:00 nmbd -D -l /var/opt/samba/bird/l
ogs -s /etc/opt/samba/smb.conf.bird
root 10114 1 0 14:44:19 ? 0:00 smbd -D -l /var/opt/samba/bird/l
ogs -s /etc/opt/samba/smb.conf.bird
root 10122 10114 0 14:44:20 ? 0:00 smbd -D -l /var/opt/samba/bird/l
ogs -s /etc/opt/samba/smb.conf.bird

So, when I stop the cluster, one smbd stops, but the other doesn't.
Is this the normal behavior or something wrong with samba.cntl and samba.mon?

[Geoff Wild]

Sounds wrong to me - when samba shuts down - all daemons should stop....

What version cntl file are you using? (it is at the top - mine in A.11.09)

Rgds...Geoff

[Oleg Podoukov]

As I found out, this is normal behavior to have 2 smbd daemons (one is a child process). Cluster stops normally, no daemons stay running.