3

I'm trying to connect to a vSphere cluster using the information from the libvirt documentation.

$ virsh -c "vpx://root@10.51.4.11/dc1/dc1-cluster-e01/dc1-vsphere-e04/?no_verify=1"
Enter root's password for 10.51.4.11: 
error: internal error HTTP response code 400 for call to 'Login'
error: failed to connect to the hypervisor

I seem to be able to establish a connection, but it fails with a "HTTP code 400". If I provide the incorrect password it fails with a 'login credentials' error, so it looks like I am getting a connection, but it's failing for another reason.

Wireshark is no help as it's all done over SSL/TLS.

Any thoughts folks?

SOLUTION: jump to my answer below about special characters.

UPDATE: 15:21 28/02/11

FYI - I'm running libvirt-0.8.3 (the Ubuntu package recompiled with the ESX flag enabled).

When I put virsh into debug mode it returns this:

[snip]
Enter root's password for 10.51.4.11: 
15:19:09.011: debug : do_open:1249 : driver 3 ESX returned ERROR
15:19:09.011: debug : virUnrefConnect:294 : unref connection 0x98aa8f8 1
15:19:09.011: debug : virReleaseConnect:249 : release connection 0x98aa8f8
error: internal error HTTP response code 400 for call to 'Login'
error: failed to connect to the hypervisor

I've watched the vmware hostd.log file whilst trying to login, and it reports this error:

UPDATE: 10:26 02/03/11

[2011-03-02 10:24:16.415 26AF8B90 verbose 'Proxysvc Req16870'] New proxy client SSL(TCP(local=10.4.100.26:34001, peer=10.51.4.103:443))
[2011-03-02 10:24:16.430 26933B90 error 'SoapAdapter'] Client sent us an invalid SOAP request: at line number 8, not well-formed (invalid token)
Coops
  • 5,967
  • 1
  • 31
  • 52

3 Answers3

2

I posted my problem on the libvirt mailing list and got a quick reply.

It turns out libvirt doesn't escape special characters in passwords currently, and as it would happen, the password had an & in it.

I can now login by using & instead of &. There is a patch in the works for it to fix this.

Coops
  • 5,967
  • 1
  • 31
  • 52
0

Run command(without last slash):

$ virsh -c "vpx://root@10.51.4.11/dc1/dc1-cluster-e01/dc1-vsphere-e04?no_verify=1"

Try:

$ virsh -c "vpx://root@10.51.4.11?no_verify=1"

or:

$ virsh -c "vpx://10.51.4.11?no_verify=1"
$ virsh -c "esx://10.51.4.11?no_verify=1"
ooshro
  • 10,874
  • 1
  • 31
  • 31
  • Nope, no luck. It still responds exactly the same. FYI, I've updated my question with some more information. – Coops Feb 28 '11 at 15:22
  • Still no luck, the shortened vpx:// version complains it needs to know the resource, and the shortened esx:// version returns HTTP 400 too. – Coops Mar 01 '11 at 10:57
0

Maybe this link will help? He's running 0.8.6 on Ubuntu...

JakeRobinson
  • 2,886
  • 17
  • 26
  • I'll try it out when I'm back in the office tomorrow, however looking it at it the article doesn't do much different from what I do already, but worth a go. – Coops Mar 01 '11 at 20:30
  • I'm thinking the version difference is enough to try it. A 400 status code seems to indicate that virsh is sending bad data, so possibly a bug? – JakeRobinson Mar 01 '11 at 20:52
  • Nope - recompiled from source using the newest version and the error returned is the same. I've updated my question with details of the vmware logs though. – Coops Mar 02 '11 at 10:31