cluster : ocfs2, gfs2 don't work ... neither gfs [message #35970] |
Thu, 07 May 2009 11:46 ![Go to next message Go to next message](/theme/ovz3/images/down.png) |
luangsay
Messages: 2 Registered: May 2009
|
Junior Member |
|
|
Hello,
I want to have my VEs on a shared volume so I have looked for a suitable file system.
I discarded the following filesystems :
* NFS isn't supported by openvz (see http://forum.openvz.org/index.php?t=msg&goto=13700&) : it has quotas problems and is very slow due to lockd issues.
* OCFS2 : vzquota don't work
* GFS2 : vzquota don't work and you've got to write a CTID.mount script to deal with a /dev issue.
So at last I tried gfs which seemed to work well (vzquotas are OK and VEs run fast).
But after creating my first real VE on gfs, VE hanged just after 10 hours.
Lot of processes are in "D state" :
[root@srvznsm proc]# ps axf -o pid,stat,wchan,cmd
PID STAT WCHAN CMD
17838 D glock_ /usr/bin/rrdtool -
17950 D glock_ php /var/www/cacti/poller.php
18004 Z exit \_ [rrdtool] <defunct>
18117 D glock_ php /var/www/cacti/poller.php
18171 D glock_ \_ /usr/bin/rrdtool -
There is no way to kill those processes and thoses processes will never die.
What is more, any other IO operation in the /vz/private/110/var/www/cacti/rra directory will end in "D state".
Looking in google, I saw that other people experienced such lock/gfs problem. Mainly with php applications.
Do anyone know how to cope with this gfs issue?
Do someone know a valid share filessystem that suits well with openvz?
Sourygna
PS : here is the versions of my RPMs (my OS is a RHEL 5.3) :
gfs-utils-0.1.18-1.el5
gfs2-utils-0.1.53-1.el5_3.2
vzctl-lib-3.0.23-1
vzrpm44-4.4.1-22.5
ovzkernel-2.6.18-92.1.18.el5.028stab060.8
vzquota-3.0.12-1
vzctl-3.0.23-1
vzpkg-2.7.0-18
vzrpm43-4.3.3-7_nonptl.6
vzyum-2.4.0-11
vzrpm43-python-4.3.3-7_nonptl.6
ovzkernel-devel-2.6.18-92.1.18.el5.028stab060.8
GFS volume is mounted this way :
[root@srvz01 ~]# grep vz01_lv /proc/mounts
/dev/mapper/vzsansun1a01_vg-vz01_lv /mnt/vz01_lv gfs rw,hostdata=jid=0:id=196609:first=1,localflocks,noquota 0 0
Now I only have one node in my "cluster" but I plan to have a second one in the future.
|
|
|
Re: cluster : ocfs2, gfs2 don't work ... neither gfs [message #36255 is a reply to message #35970] |
Tue, 02 June 2009 07:02 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) |
luangsay
Messages: 2 Registered: May 2009
|
Junior Member |
|
|
Seems like new kernel 2.6.18-128.1.1.el5.028stab062.3 solved that issue.
I've been using it for two weeks, with two VE whose simfs filesystem lies on a gfs filesystem and I haven't seen the previous lock problem.
What's more, I've seen that another issue of gfs filesystem had been solved : now we can mount it with noatime and nodiratime options (before, we had the error "error mounting ... Invalid argument).
Sourygna
|
|
|