[Linux-cluster] GFS 6.0u5 not freeing locks?

Wendy Cheng wcheng at redhat.com
Wed Dec 14 06:03:31 UTC 2005


On Tue, 2005-12-13 at 20:12 -0600, vze8dkpl at verizon.net wrote:
  
>   
> Thanks for the response.  The machines I am using for testing HP 
> DL380 w/2 cpu's and 3GB of ram. Can you explain why GFS stopped 
> cleaning up after itself even though the high water mark was set to 
> 10000? The locks remained at 300K+ and never went back down. 

This is first time we have such report (lock count higher than
lt_high_locks) since normally we have users set "lt_high_locks" to
higher values than default 1024*1024. Will look into this issue more.
Intuitively I suspect there is a time lag between sending and receiving
of "lock full" messages between daemons

>  
> I'd be glad to try the new revision, just let me know how to get 
> it. Also a list of tunables would be good as well. 

Is it possible for you to log a support call or open a bugzilla (so we
can have some records internally) ? Will work with our maintainer
tomorrow to put the test RPMs somewhere that you can retrieve. 


-- Wendy




More information about the Linux-cluster mailing list