Samba process using 100% of CPU time, New SAMBA$SMBD_startup.log file generated every minute.

[Simon Hutchings_1]

I have inherited a Samba/CIFs configuration that is running on RX3600 IA64 server, it seems to be using most of the processor time to service the application.
Are my observation typical for a Samba configuration ?
How offend should the system generate and new SAMBA$SMBD_startup.log file ?
I am seeing a new file being generated about every minute.
The Rx3600 is a 4 processor systems which quite offend can have 2 or 3 SMBD445 processes running taking 99% of the CPU time per processor.
0 25 50 75 100
+ - - - - + - - - - + - - - - + - - - - +
001C2463 SMBD445_BG15349 99 #####################
001B245B SMBD445_BG15303 99 ####################
001A2862 SMBD445_BG15343 98 ###################
0000044E DESTA Director 1
Despite these observation I am getting no completes from users about the remotes share.
[global]
workgroup = VISTA
server string = Vista Development Server
netbios name = %h
security = user
passdb backend = tdbsam
map to guest = Bad User
log file = /samba$root/var/log_%h.%m
domain logons = Yes
guest account = SAMBA$GUEST
domain master = no
create mask = 0755
load printers = no
printing = OpenVMS
guest ok = Yes
debug level = 6
idmap uid = 10000-20000
idmap gid = 10000-20000
printcap name = /dev/null

[homes]
comment = Home Directories
browseable = no
read only = no
create mode = 0750
vms ods5 volume = Yes

[Data]
comment = Data
path = dsa3:[data]
read only = No
vfs objects = varvfc
vms ods5 volume = Yes

[End of file]

[Shilpa K]

Hi Simon,

From the SMB.CONF parameters, it appears like you have configured CIFS as Backup Domain Controller (BDC). Do you have a Primary Domain Controller (PDC) for the domain VISTA? From the parameters "map to guest" and "guest ok", it appears like you want to allow any user to connect to CIFS Server. I suggest you to contact HP so that based on the requirements at your site, a correct CIFS Server configuration can be suggested to you.

Regards,
Shilpa

[Rick Retterer]

Is your setting for "debug level" too high?

What types of errors are you seeing in the log files for Samba?

We need more information to be able to provide you with a thorough answer.


Rick

[Shriniketan Bhagwat]

Hi,

I do not know about the SAMBA and my below update may not be relevant for this query, But it is for the users to limit the CPU usage. Thought it may be useful for some one.

OpenVMS has a feature called CLASS SCHEDULING. Class scheduling is implemented in the SYSMAN utility. The class scheduler gives the ability to limit the amount of CPU time that a system's users may receive by placing the users into scheduling classes. Each class is assigned a percentage of the overall system's CPU time. As the system runs, the combined set of users in a class is limited to the percentage of CPU execution time allocated to their class.

Class scheduling can be used when the application consumes entire CPU processing power and causes other users in the system to wait. In this case add the user to the class who runs such applications by specifying the limited percent of CPU. This will limit the application from consuming the entire CPU.

Refer HP OpenVMS System Management Utilities Reference Manual:M-Z for more information on class scheduling.
http://h71000.www7.hp.com/doc/os83_index.html


Regards,
Ketan

[Paul Nunez · Accepted solution]

Hi Simon,

It's not normal for any smbd process to consume 99% of a cpu. You can determine which clients are associated with these processes by taking the BG device referenced in the process name and doing:

$ ucx sh dev

Also $ smbstatus will show what shares those users are mapped to, which may provide a clue as to where the problem lies.

But once you know the name of the workstations of these processes, you can check samba$root:[var]log_., as well as old log files (extention ends in _old) for clues. With "log level = 6" I suspect the logs will help.

And as Shilpa noted, the server is configured as a Backup Domain Controller in the Vista domain, probably unintended.

Paul

[Craig A Berry]

You also need to have a look at what the clients are doing. See Jim Duff's blog post entitled "Polling Samba":

http://www.eight-cubed.com/blog/archives/001178.html#001178

The CPU load he describes sounds similar to yours, though of course there could be different reasons.