Just playing with Corosync so I thought I would include a simple howto. For examples sake, we will setup an OpenVPN server with a fail-over.
I set this up on 2 KVM guests using the default network configuration. In the examples I will use the ip addresses of my guests, change them to match your setup.
Continue reading Centos 6, Corosync and Pacemaker for a simple Active/Passive cluster
I recently had a problem where a yum update broke things because of my umask setting, I needed to reinstall the packages.
Yum doesn’t have a reinstall feature, and I didn’t want to do a remove first, the packages were currently in use. So how to do a reinstall?
Remove the package from the RPM database without actually removing it:
# rpm -e --justdb --nodeps "package name"
The package is still installed but when yum queries the RPM DB it will find it missing and allow you to install it:
# yum install "package name"
I really should read man pages. I had no idea that rpm could do rollbacks with just a couple of config changes, and that it was supported in yum.
This would have come in really handy last week. I upgraded an out of date file server and the latest samba broke Directory authentication, which was not cool and file shares cold not be authenticated until the problem was solved. CentOS do not keep old packages in their repos. I had to compile a FC6 version of samba to rollback to a working state until I investigated and found a patch.
The bug is dated May, the patch has been out for months, and the fix is still not in official repos.
Anyway back to rollbacks, 2 steps:
- In /etc/yum.conf add the line:
- In /etc/rpm/macos (create it if it does not exist), add the line:
Now if the update or install clobbers something, you can roll back packages to their former state just with a simple rpm command.
rpm -Uhv –rollback ‘9:00 am’,
rpm -Uhv –rollback ‘4 hours ago’,
rpm -Uhv –rollback ‘december 25′.
Wow déjà vu. So gnome-vfs is still brain dead. Of course NFS4 is going to be used at some point for home directories and need trash support. Why is the list of supported filesystems still hard-coded in a c file?
Follow the instructions from the previous post, modify them to say nfs4 instead of fuse.
Patch file you can apply to the SRPM of gnome-vfs: gnome-vfs-2162-nfs4-trash.patch
My latest lab build is using nfs4 to mount home directories.
Kerberos authentication requires the time to synced before it will work. Redhat (CentOS) tries mounting Network filesystems before syncing the time on startup. Can anyone else see the problem?
If any of our lab machines have their time out by more than 5 minutes the mount will fail.
I have changed the startup priority of ntpd from 58 to 12, and this has fixed the problem.
As ntpd only requires networking to work, I cannot understand why it starts at 58, but problem solved.
I have been setting up an Xdmx render wall. When finished it will be a 3×3 rear projected screen, with 9 computers (CentOS 5 installs) and 18 projectors for 3d visualization.
For a couple of years now Xdmx has been rolled into the Xorg source and getting it configured is really not that hard. When I have ironed out this install I may write a howto. But while some stuff was working just fine and dandy most apps that use a modern widget library for the UI (ala GTK or QT) were failing to render fonts properly, not drawing icons, messing up colors and just not looking right in general.
I scoured the interweb for answers, and the main thing I saw was just forum posts instructing users to turn of the Xorg render extension. This works, though makes the rendering of windows very, very slow. On what should be a high performance render wall I have an issue with this.
Anyway, I eventually found this post, and figured it was worth a try. And it Worked! brilliant! QT, GTK, Firefox, everything is looking peachy.
I threw it together in this patch file which I applied to the SRPM for xorg-x11-server.
huh? I can’t seem to find pam_mount for CentOS 5. RPMForge seems to have the build logs for it, but yum can’t see the rpm to install and neither can I.
Pretty simple to fix, a bit lazy, but it will only take you 5 minutes.
Grab the rpm sources for libHX and pam_mount from your local Fedora 9 mirror. I got mine from here:
Build the libHX source:
# rpmbuild --rebuild libHX-1.18-1.fc9.src.rpm
Install the libHX rpm and the libHX-devel rpm. Now build pam_mount the same way:
# rpmbuild --rebuild pam_mount-0.41-2.fc9.src.rpm
Use “yum localinstall” to install pam_mount as it probably needs to grab a perl XML module as well.
Leopard has a nifty VNC implementation built right into the OS, you can share your screen from the Sharing preferences and connect to it using any VNC client. And there is a VNC client built right into the finder ⌘-K (Connect to server) allows you to enter in a vnc server in the form vnc://server:port.
But if you try and connect to an up to date Xvnc server it will probably fail, you need to specify that your VNC server uses the VNC protocol 3.3 for backwards compatibility, the command option is simply “-Protocol3.3” you will now be able to connect to your Xvnc server from Leopard using the built in viewer.
If you were to follow my earlier instruction for a GDM VNC Server then simply add this option to: /etc/xinetd.d/vnc
You will notice that the connecting to server box doesn’t go away, this is because Apple does not like the option: -securitytypes=none
Setup a password file using vncpasswd and remove the above option and add the option:
It is possible to allow remote users to login to gnome, using gdm as you would on console, via vnc. There are other tools to do this, like NX, which are technically superior, but vnc clients are almost ubiquitous making this solution work for almost any client.
Continue reading GDM logins using VNC