Date: Sat, 30 Oct 2004 13:20:08 -0300 (ADT) From: "Marc G. Fournier" <scrappy@hub.org> To: freebsd-stable@freebsd.org Subject: vnode 'leak' in 4.x ... Message-ID: <20041030131002.O6085@ganymede.hub.org>
next in thread | raw e-mail | index | archive | help
A little while ago, I reported a suspicion that vnodes just weren't being freed up on long running servers ... after 55days of uptime on one of my servers, here is what I'm dealing with ... 793 'samples' today (one every minute) 786 with vnlru in a vlrup state I shutdown all of the VMs running on the large hard drive (the only place unionfs is being used) and umount'd the drive ... there were some suggested back then that this might/should free everything back up again ... but it didn't: Oct 30 13:06:02 venus root: debug.numvnodes: 522265 - debug.freevnodes: 57966 - debug.vnlru_nowhere: 209679 - vlruwt Oct 30 13:07:00 venus root: debug.numvnodes: 522265 - debug.freevnodes: 57268 - debug.vnlru_nowhere: 209679 - vlruwt Oct 30 13:08:00 venus root: debug.numvnodes: 522265 - debug.freevnodes: 52335 - debug.vnlru_nowhere: 209679 - vlruwt Oct 30 13:09:00 venus root: debug.numvnodes: 522265 - debug.freevnodes: 50228 - debug.vnlru_nowhere: 209682 - vlrup Oct 30 13:10:01 venus root: debug.numvnodes: 522265 - debug.freevnodes: 44407 - debug.vnlru_nowhere: 209690 - vlrup Oct 30 13:11:00 venus root: debug.numvnodes: 522265 - debug.freevnodes: 35424 - debug.vnlru_nowhere: 209697 - vlrup Oct 30 13:12:02 venus root: debug.numvnodes: 522265 - debug.freevnodes: 34626 - debug.vnlru_nowhere: 209708 - vlrup Oct 30 13:13:00 venus root: debug.numvnodes: 522265 - debug.freevnodes: 29214 - debug.vnlru_nowhere: 209727 - vlrup Oct 30 13:14:00 venus root: debug.numvnodes: 522265 - debug.freevnodes: 24414 - debug.vnlru_nowhere: 209746 - vlrup Oct 30 13:15:00 venus root: debug.numvnodes: 522265 - debug.freevnodes: 26994 - debug.vnlru_nowhere: 209766 - vlrup The 'vlruwt' states above are while I had everything shutdown ... the vlrup's all started again after I mounted the drive and started to restart the VMs themselves ... I expect a high # of vnodes to be used ... that isn't the issue ... the issue is that even getting rid of the major mount point, so that only /, /tmp, /usr, /var are left up, the large # of vnodes that are in use on that mount point aren't being freed by vnlru :( I hate to reboot the server, but it looks like I've got no choice at this point ... is there something else that I can do, in 50 days or so, to provide more information? Thanks ... ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041030131002.O6085>