kadmin incremental propagation full resync multiple processes spawned

[Paul B. Henson]

On 11/4/2011 1:40 PM, Tom Yu wrote:

> I think it should be possible to do something using alarm() to provide
> timeouts for lock acquisition, but that wouldn't work well inside a
> library.

As I understand it, the issue is that there is no "get a lock, wait in 
line if it's not available" call. If such a call did exist, it would 
need a timeout value to limit how long it would wait. But the current 
call simply returns immediately, either with the lock, or without it. If 
you didn't get it, your only choice is to try and get it again. However, 
if it does become free, there's nothing keeping someone else from 
jumping in and grabbing it before you have the chance to try again. The 
current approach tries five times with a 1 second delay in between tries 
to get the lock, then gives up. If somebody else (in this case kadmind) 
is also busily grabbing the lock, and sitting on it, the odds of you 
getting in to get it aren't necessarily that good.

Barring a better lock API, the only other little choice is to try more 
times and perhaps wait less in between tries. I suppose you could spin 
on it, continuously trying to get the lock for a certain interval, but 
that would be a bit inefficient.


-- 
Paul B. Henson  |  (909) 979-6361  |  http://www.csupomona.edu/~henson/
Operating Systems and Network Analyst  |  henson at csupomona.edu
California State Polytechnic University  |  Pomona CA 91768