[Linux-cluster] DLM behavior after lockspace recovery

[Daniel Phillips]

On Friday 15 October 2004 10:20, David Teigland wrote:
> On Fri, Oct 15, 2004 at 12:32:38AM -0400, Daniel Phillips wrote:
> > But do you really think the dlm should pretend that a potentially
> > corrupt value is in fact good?  This seems like a very bad idea to
> > me. In return for saving some bookkeeping in the very special case
> > where you have an incrementing lvb, you suggest imposing extra
> > overhead on every lvb update and having the dlm make false promises
> > about data integrity. I don't think this is a good trade.
>
> Incorrect.  Nothing is corrupt, there's no "false promise", there's
> no overhead to speak of, and restoring the last available value is
> standard behavior.

A standard bug you mean.  Suppose, in my application, I were to store 
the number of nodes currently writing to a disk region in a lvb.  Each 
time a node initiates a write it gets PW and increments the value.  
When the write completes, it decrements the value.  If zero, the region 
is marked as "in sync".

Now the lock master dies and an older lvb value is quietly substituted 
for the actual value.  You see that a region is now going to be marked 
as "in sync" when there are still one or more writes outstanding to it.

The correct behavior is obviously to mark the lvb invalid and have the 
application respond by contacting each node to see which of them are 
still writing to the region.

I hope you see why it is in general, bad to lie about the integrity of 
data.

Regards,

Daniel