Andy Shellam's questions

I've been trying to set up a pair of servers utilising KVM on Ubuntu 9.10 to host 8 virtual machines between them and ended up with various issues from the VMs freezing, to not powering on.

I had one virtual server set up and running and was setting up a second, when any operation involving OpenSSL would cause the VM to lock up in a weird way - all network traffic would cease, it wouldn't process logins on the console, but it wasn't taking any CPU time off the host. The first virtual server was identical and worked perfectly.

Another VM I tried to setup had installed Ubuntu fine then refused to reboot, throwing kernel exceptions to do with XFS.

I've now installed Citrix XenServer 5.5 on both hosts, and am now setting up my third VM with absolutely no issues. I also had the same experience when I tried VMware, but I preferred Xen as it appears to give more features on the free license.

My question is am I just unlucky with KVM, or is KVM as unstable as it appears? Are you using, or planning on using, KVM in production, and how successful have you been?

EDIT:

Thanks for all the answers. I have recently upgraded our two physical servers to Ubuntu 10.04 LTS which uses KVM 0.12.3. I'm now installing my sixth virtual machine on KVM and haven't had even a hint of the issues I experienced previously, so I'm now sticking with KVM :)

I have 2 identical servers running Ubuntu 9.10 and I've been playing with Ubuntu's upstart system, trying to write a sort-of complex boot order for a few services.

I've also played with init scripts to try and achieve the required effect, but I've now removed all my initscripts.

I've been doing this only on one server. Part way through my development of my upstart scripts, I noticed that the server wasn't rebooting properly - it kills off rsyslog-kmsg then just sits there. If you run the reboot command again, it reboots.

I removed all traces of the scripts I created and still it won't reboot, so I'm now thinking of done something but I cannot work out what I've done that I haven't reversed.

The other server is perfectly fine. I've compared the timestamp of the upstart scripts (in /etc/init) and copied over the init.d, rc[0-6].d directory contents onto the first server, but it still won't reboot.

The strange thing is, if you do halt instead of reboot, it powers off as normal - it just won't reboot.

There are no messages in dmesg or /var/log/messages after the boot process completes.

Any ideas? I'd prefer not to have to re-install the OS as it's a remotely-hosted server.

I was hoping some PAM/LDAP gurus may be able to help me out here. I've recently set up an LDAP directory on Ubuntu Server to hold accounts for both my clients (for use with web-based systems) and staff (who will need to login via SSH.)

The LDAP authentication is working perfectly. However I cannot get the account restrictions working: staff accounts will have IDs between 2001 and 2999 and will be a member of the ssh-users group to allow them to login to servers.

The restrictions in question are in /etc/ldap.conf, and are pam_min_uid, pam_max_uid and pam_groupdn.

pam_groupdn contains the full DN to my ssh-users group. pam_min_uid = 2000 and pam_max_uid = 2999.

Now, I've managed to get them working by adding:

account [success=1 default=ignore] pam_ldap.so

above the pam_unix.so line in /etc/pam.d/common-account. However, the local Unix accounts can then NOT login: the SSH server kills the connection as soon as they try.

I've set the pam_ldap.so module to sufficient in the above file, but then the invalid users get a message saying they cannot login, but it logs them in anyway.

So, how can I set these account restrictions for LDAP users, while still allowing UNIX users to login?

As you can probably guess I'm a newbie to PAM, although I have managed to get the "automatically make home directories" module working :-)

Many thanks, Andy