WARNING: Major vnode leakage in RELENG_4 NFS ...

Marc G. Fournier scrappy at hub.org
Mon Jul 18 04:38:32 GMT 2005


Hopefully that caught enough attention :)

I've recently been having problems with one of my servers *hanging* that 
I've just recently traced down to what looks to be a vnode leak in the 
latest NFS code in the RELENG_4 branch ...

Unfortunately, my 'debugging' is at a stand still, but if anyone is 
noticing issues with their 4.x, where the server seems to hang, take a 
look at your 'sysctl debug.freevnodes' while doing some operations on an 
NFS mount ...

To "compensate" for the issue on my server, I've changed my scripts so 
that they umount;mount the NFS file system before doing any major 
operations, to free up the vnodes, which so far seems to be preventing a 
reoccurance of the lock up ...

As an example, I have a process that starts at 1am that does a find across 
some NFS mounted directories, and my freevnodes has drop'd by ~130k:

Jul 18 01:00:00 mercury root: debug.numvnodes: 460354 - debug.freevnodes: 431364 - debug.vnlru_nowhere: 0 - vlruwt
Jul 18 01:12:00 mercury root: debug.numvnodes: 460354 - debug.freevnodes: 355260 - debug.vnlru_nowhere: 0 - vlruwt
Jul 18 01:24:00 mercury root: debug.numvnodes: 460354 - debug.freevnodes: 307395 - debug.vnlru_nowhere: 0 - vlruwt

Without the umount;mount, they drop farther/faster :(




----
Marc G. Fournier           Hub.Org Networking Services (http://www.hub.org)
Email: scrappy at hub.org           Yahoo!: yscrappy              ICQ: 7615664


More information about the freebsd-questions mailing list