RedHat Advanced Server 3
kernel = 2.4.21-15.0.4.ELsmp
samba-3.0.4-6.3E
scsi hdd's

I setup a dell server with samba as a primary domain controler. All the services run well. I can create shares and transfer files like one would expect from a file server. My problem arrises when I map to my "ghost" share and attempt to reghost a machine. The ghost session quickly fails after it begins with some error reported by ghost (sorry, dont have the message to write down).

The server (using dmesg) reports "lease broken - owner pid = 3589". Actually, any large file transfer generates this kernel error. Top shows that there is high cpu access and very high iowaite cycles.

I attempted to tweak the /etc/sysctl.conf file with

vm.bdflush = 90 2000 0 0 5000 10000 100 50 0
vm.kswapd = 1024 64 16
vm.page-cluster = 5
vm.inactive_clean_percent = 100
vm.pagecache = 2 10 20

but the same errors keep showing up.

Oplocks in samba make no difference either turned on or off.

I've automounted the hdd that I use as my share target in fstab with "/dev/sdb1 /mnt/sdb1 ext3 defaults 0 1". I'm not sure if my options are correct or not though...

I'm pretty stumped as to what I can do to fix this problem. Anyone have any ideas?

Here is an update incase anyone is trying very hard to figure out my problem...

My laptop (Thinkpad T41) is running the same version of redhat as my Dell server (mentioned above). All the same updates are applied etc...
The only difference between the server and my laptop is hardware.

I have tried to duplicate the problem that I'm having with the server. By the looks of things my laptop is flawlessly running. I am not getting any "lease broken" kernel messages with the thinkpad. My only guess then is that my servers hardware (scsi drive?) is not playing nicely with the kernel/whatever.

Are there any kernel params that I can pass at boot up for the scsi that would improve performance? My problem is a bit too vague to find any answers on the net...

thanks

update to this update :: I tried running a ghost session from my laptop and it failed as well. There are no broken oplocks listed or broken leases etc... in any logs or any other location. Its much stranger than I thought. The server and laptop have different hardware. One is scsi the other is ide. Bus type does not seem to be an issue.

Kernel oplock failure? Oplocks are buggy in the kernel, but this bad?

thanks again.

Dar