[Samba] Stale byte range locks - How can I remove them w/o restart?

John Laur john at energynet.com
Tue Mar 29 00:20:47 GMT 2005


I am having a problem that I have seen described in various bug reports..
One filed bug (#1267) was supposed to have been fixed in 3.0.7 but this
problem is still existing in 3.0.10 from what I can tell. Here's what I see:

mercury:/root# smbstatus -LB
----------8<----------<SNIP>----------8<----------
Byte range locks:
   Pid     dev:inode  R/W      start        size
------------------------------------------------
  3075   00901:4e8ce    W        563           1
  3075   00901:23eb30   W        563           1
mercury:/root# ps auxw | grep 3075
root      6506  0.0  0.0  1340  472 pts/1    S    16:44   0:00 grep 3075
mercury:/root# smbd -V
Version 3.0.10-Debian
mercury:/root#

The application is quickbooks. Every once in a while, a user will get some
kind of generic and non-specific "network error" message in Quickbooks and
be kicked out of the application. Apparently the smbd serving their requests
dies. While this may indicate an entirely different problem, the main issue
at hand is that stale byte range locks are left by the dead smbd and I
cannot clear them without completely restarting samba.

Is there a workaround until this bug gets fixed or if it is possibly fixed
in a later samba release is there a workaround or some other way to purge
these stale locks until I can do some QA between our apps and samba before
rolling out a new release or resorting to restarting smbd?

Thanks to anyone who can help!
John Laur



More information about the samba mailing list