| ovzkernel 2.6.16-026test009.1 and RHEL 4 update 3 problem [message #2786] |
Thu, 20 April 2006 19:51  |
 |
dowdle
Messages: 261
Registered: December 2005
Location: Bozeman, Montana
|
Senior Member |
|
|
Greetings,
I installed ovzkernel 2.6.16-026test009.1 on two machines today. One was running CentOS 4.3 and the other Red Hat Enterprise Linux 4 Update 3... completely different hardware.
Anyway...
Here's the problem I'm having on the RHEL machine.
1) Apache is in a VPS
2) The physical host has several NFS mounts
3) The NFS mounts are bound to the VPS
4) Apache in the VPS has one of the NFS mounts as DocumentRoot
5) The web pages are serving the text portions just fine but embedded .jpg's in pages just sit and spin... and eventually time out. I can reload the page and get some of the images to show up.
6) Nothing looks out of the ordinary in /var/log/httpd/*
7) Nothing looks out of the ordinary in the physical host logs
8] /proc/user_beancounters is all 0's
On the first attempt to install the test kernel, I got an rpm conflict with a package named:
lksctp-tools
I removed that package and the test kernel installed just fine.
Any idea what is going on? I've restarted apache and the VPS but that didn't help. NFS is working fine... and the files are available in the VPS... apache just is not transferring them... or the HTTP protocol is getting confused or something.
All other services appear to be working fine so I'm assuming it isn't some sort of connectivity issue. iptables is NOT running on the physical host at this time.
Any ideas?
I did not get the package conflict when I installed the test kernel on CentOS 4.3 and oddly enough the RHEL install was a minimal install... so the package I had to remove must have been a default?!? Of course, like I said... the CentOS is completely different hardware and isn't doing the NFS stuff at all.
--
TYL, Scott Dowdle
Belgrade, Montana, USA
|
|
|
|
|
|
| Local files, no problem: NFS files, problems [message #2788 is a reply to message #2787] |
Thu, 20 April 2006 22:03  |
|
dowdle
Messages: 261
Registered: December 2005
Location: Bozeman, Montana
|
Senior Member |
|
|
I used wget and dragged some of the files over into a local directory and then modified Apache's DocumentRoot... restarted apache... and no problem at all.
I switch it back, and problems.
The same configuration did not give me a problem with the stable OpenVZ kernel.
Another question... sorta related but not.
Ok, in /etc/sysconfig/vz-scripts I have 101.mount and 101.umount and both are 755.
Check out the following behavior:
(before starting 101 - physical host)
[root@hprack6 ~]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/cciss/c0d0p1 30G 1.8G 27G 7% /
none 1012M 0 1012M 0% /dev/shm
esus:/students 145G 31G 107G 23% /mnt/nfs/esus/students
esus:/home 145G 112G 26G 82% /mnt/nfs/esus/home
esus:/www/htdocs 55G 11G 43G 20% /mnt/nfs/esus/www
[root@hprack6 ~]#
(after starting vps - physical host)
[root@hprack6 ~]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/cciss/c0d0p1 30G 1.8G 27G 7% /
none 1012M 0 1012M 0% /dev/shm
esus:/students 145G 31G 107G 23% /mnt/nfs/esus/students
esus:/home 145G 112G 26G 82% /mnt/nfs/esus/home
esus:/www/htdocs 55G 11G 43G 20% /mnt/nfs/esus/www
/mnt/nfs/esus/home 145G 112G 26G 82% /vz/root/101/home
/mnt/nfs/esus/students
145G 31G 107G 23% /vz/root/101/students
/mnt/nfs/esus/www 55G 11G 43G 20% /vz/root/101/www
[root@hprack6 ~]#
(after starting the vps - vps)
[root@web2 ~]# df -h
Filesystem Size Used Avail Use% Mounted on
simfs 5.0G 426M 4.6G 9% /
nfs 145G 112G 26G 82% /home
nfs 145G 31G 107G 23% /students
nfs 55G 11G 43G 20% /www
[root@web2 ~]#
(after stopping vps - physical host)
[root@hprack6 ~]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/cciss/c0d0p1 30G 1.8G 27G 7% /
none 1012M 0 1012M 0% /dev/shm
esus:/students 30G 1.8G 27G 7% /mnt/nfs/esus/students
esus:/home 30G 1.8G 27G 7% /mnt/nfs/esus/home
esus:/www/htdocs 30G 1.8G 27G 7% /mnt/nfs/esus/www
[root@hprack6 ~]#
Notice how all of info for the nfs mounts has chaanged and oddly enough they reflect the / filesystem rather than their real values? CDing into those dirs and doing an ls shows NO files... everything has disappeared. When I umount the dirs by hand they say they aren't mounted... yet they disappear from the df listing... and then I mount -a and they get mounted correctly again.
Here are the contents of my 101.mount script:
#!/bin/bash
#
# 101.mount - script to bind nfs mounts of physical host
# if one of these files does not exist then something is
# really broken
[ -f "/etc/sysconfig/vz-scripts/101.conf" ] || exit 1
# If home dirs are not mounted we exit with error
mount --bind /mnt/nfs/esus/home /vz/root/101/home
mount --bind /mnt/nfs/esus/students /vz/root/101/students
mount --bind /mnt/nfs/esus/www /vz/root/101/www
exit $?
Here are the contents of my 101.umount script:
#!/bin/bash
# if one of these files does not exist then something is really broken
[ -f "/etc/sysconfig/vz-scripts/101.conf" ] || exit 1
# umount nfs mounts
umount /vz/root/101/home
umount /vz/root/101/students
umount /vz/root/101/www
exit $?
Any idea what is going on? Sorry for the bad formatting. The <pre> tag doesn't seem to work for me here.
I'm not sure if this is related to the test kernel or if I had the same problem with the stable kernel. I installed the new vzctl with the migration features... which requires the test kernel. If I reboot and run the stable kernel am I in need of downgrading vzctl anv vzctl-libs?
--
TYL, Scott Dowdle
Belgrade, Montana, USA
[Updated on: Thu, 20 April 2006 22:28] Report message to a moderator |
|
|
|
|
|
|
|
|
|
|
|
| Re: ovzkernel 2.6.16-026test009.1 and RHEL 4 update 3 problem [message #2884 is a reply to message #2827] |
Mon, 24 April 2006 22:25 |
|
dowdle
Messages: 261
Registered: December 2005
Location: Bozeman, Montana
|
Senior Member |
|
|
Ok, I uncommented the following line in the apache config:
EnableSendfile off
I wasn't familiar with that option until you brought it to my attention. I restarted apache and that didn't seem to make a difference.
I checked the status of the running apache processes and none of them were in a D state... they were all sleeping.
I'm not that familiar with using strace... so I didn't get anything useful... and it seemed to detach almost immediately.
I tailed the access_log while browsing pages... sometimes the pages would show as 200 (sent) but often my browser didn't get them... somtimes the request was never recieved by the server.
I had a kernel Opps and had to reboot. There was nothing intelligable in the opps message. I decided to boot to the stable kernel.
I did double check that apache in a vps has no problem serving up the files from the stable kernel... but oddly enough the weird behavior with NFS mounts *IS* there with the stable kernel.
1) on physical host, mounts look fine
2) stop my vps
3) all of the mounts on the physical host reflect the info of the / filesystem
4) I have to umount all nfs mounts and mount them again for them to work again
Weird.
--
TYL, Scott Dowdle
Belgrade, Montana, USA
|
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
