[libvirt] ESXi support

[Matthias Bolte]

2010/7/3 adrian wyssen <wyssenadrian at gmail.com>:
> Hello everyone
> I am really new to libvirt and I have to use it with ESXi. I was writing
> code for listing the virtual machines in a ESXi server. If I connect me with
> the virsh to the ESXi server, it works. But since I can not create any
> certificates on the server (ESXi does not have any console...) I have to use
> the no_verify option (esx://host/?no_verify=1)
> I want to do the same with the API, but when I use this string in the uri
> parameter of the following code
> conn = virConnectOpenAuth(uri, virConnectAuthPtrDefault, 0);
> the system returns "libvir: Remote error : Cannot access CA certificate
> '/etc/pki/CA/cacert.pem': No such file or directory
> No connection to hypervisor"
> What method can I use to be able to connect to the server by ignoring the
> missing certificate? Another question is: Are there any example codes for
> using ESX(i), because as I said I am really new with libvirt and I did not
> find any examples.
> Thanks a lot for your help.
> Adrian

The certificates libvirt is complaining about have nothing to do with
the certificates for ESX(i).

By default libvirt uses a TLS connection to a remote libvirtd (a
daemon). This daemon is used to allow remote management of hypervisors
that don't have remote capabilities on their own, like QEMU. For
ESX(i) this daemon is not necessary, because libvirt directly uses the
remote SOAP API provided by the ESX(i) server.

By default libvirt uses a HTTPS connection to the ESX(i) server.
Proper HTTPS requires correctly configured SSL certificates, this
certificates are different from the certificates for libvirt's TLS
connection to libvirtd. In case of a default ESX(i) installation the
ESX(i) server has auto-generated self-signed certificates. And in
general those cannot be verified by your client because your client
doesn't have the cacert.pem that corresponds to the private key that
the ESX(i) server used to sign the auto-generated self-signed
certificates.

That's the reason why libvirt allows you to disable those certificate
check by appending ?no_verify=1 to the URI.

libvirt complains here about the for the TLS connection to a remote
libvirtd. This is a bit cryptic error that appears because libvirt
don't know what to do with the URI you gave to it and as last option
it tries to connect to a libvirtd on the remote server hoping that the
libvirtd there knows what to do with the URI.

The typical reason why this happens is that your libvirt is build
without ESX support. But you said 'virsh -c esx://host/?no_verify=1'.
Therefore, I assume that's not the case here.

Could it be that you have multiple version of libvirt installed? One
that has ESX support enabled and is used by virsh and one that has ESX
support disabled and that your program links against?

Regarding example code: There is no special ESX libvirt example code,
because libvirt is designed to provide hypervisor independent API.

You can take a look at the examples directory. For example the
hellolibvirt example:

   ./examples/hellolibvirt/hellolibvirt esx://host/?no_verify=1

If that still gives you that certificate error, try:

   LIBVIRT_DEBUG=1 LIBVIRT_LOG_OUTPUTS=1:stderr
./examples/hellolibvirt/hellolibvirt esx://host/?no_verify=1

It should list lines like

13:13:23.189: debug : virRegisterDriver:927 : registering ESX as driver 5
...
13:13:23.189: debug : do_open:1242 : trying driver 5 (ESX) ...

If there is no virRegisterDriver or do_open line regarding ESX then
your libvirt is build without ESX support. But i wonder why virsh
works then.

Recent (unreleased) libvirt has better error reporting for this
situation and tells you when you try to use an URI scheme that was
disabled, instead of reporting unexpected certificate errors.

Matthias