Re: Make vnc bind to 0.0.0.0 insted of localhost on XenServer 6 with CS 3.0.2

Tue, 06 Nov 2012 01:36:42 -0800 

Hi Rajesh and others,
i decided that we don't have time to discuss this anymore, so i just hacked myself qemu-dm-wrapper file. 
I used this commit as a guideline:
https://github.com/xen-org/xen-api/commit/682f85fadff8cf199c8bf932854c5ef1a9a3e0a7

this is the change i did:
[root@x1 libexec]# diff /root/qemu-dm-wrapper qemu-dm-wrapper
98a99,101
>     qemu_args.append("-vnc")
>     qemu_args.append("0.0.0.0:1")
>
After migrating one Windows VM to x1 hipervisor host, i get a working console using CS 3.0.2 GUI.
If this is somehow wrong or will have some dire consequences please let me know, otherwise i'll just do it on all hypervisors. 

Regards,
F.

On 6/11/12 10:17 AM, France wrote:

Hi Rajesh,

stopping and starting (not restarting) did not help.
The console still isn't accessible from CloudStack GUI.

On unrelated note, it works from XenCenter.
As for the wiki post, i cannot read it. I get this error upon visiting with registered username: 
-
You cannot view this page due to inherited restrictions
Page level restrictions have been applied to a parent of the current page. These restrictions limit access to certain user(s) or group(s) and apply to all pages underneath the parent. 
-

Regards,
F.

On 6/11/12 6:35 AM, Rajesh Battala wrote:

Hi France,
Moving vnc term to listen on localhost is done because of this http://wiki.cloudstack.org/display/RelOps/Secure+Console+Access+on+XenServer ( sorry to point to cloudstack wiki as I didn't find the same page in apache Cwiki).
Can you please stop the vm and then start instead of restarting the vm and let us know the issue is resolved or not? 

Thanks
Rajesh Battala



From: France [mailto:[email protected]]
Sent: Monday, November 05, 2012 8:25 PM
To: Rajesh Battala
Cc: [email protected]
Subject: Re: Make vnc bind to 0.0.0.0 insted of localhost on XenServer 6 with CS 3.0.2 

Hi Rajesh,

thank you for taking the time and helping us with this issue.
Below is additional information.
Some of the consoles started working after rebooting them via CloudStack GUI, while some (Windows) are still inaccessible. This seems to be now a Windows only issue or an issue related to: qemu-dm-wrapper script.
Currently on none of the running/old (not rebooted) instances console is available. After rebooting "Console Proxy VM" itself, i can access following system VMs: 
"Secondary Storage VM"
"Console Proxy VM"
All rebooted system routers.
If i reboot Linux based VM, router or user's instance, the console starts working again. 
If i create a new centos 6.3 VM, i also get a working web console.
If i create a new windows 2008R2 vm or if a reboot an already created one, i cannot get the console. I could before installing hotfixes. 

This is how netstat looks if for that windows VM:
tcp 0 0 127.0.0.1:5901 0.0.0.0:* LISTEN 8232/vncterm tcp 0 0 127.0.0.1:9501 0.0.0.0:* LISTEN 8232/vncterm This is how netstat looks if a live migrate another linux to the same hipervisor: tcp 0 0 127.0.0.1:5901 0.0.0.0:* LISTEN 8232/vncterm tcp 0 0 0.0.0.0:5903 0.0.0.0:* LISTEN 5065/vncterm tcp 0 0 127.0.0.1:9501 0.0.0.0:* LISTEN 8232/vncterm tcp 0 0 0.0.0.0:9502 0.0.0.0:* LISTEN 5065/vncterm 

Some additional details:
[root@x1 ~]# netstat -paltn | grep 59
tcp 0 0 127.0.0.1:5901 0.0.0.0:* LISTEN 8232/vncterm tcp 0 0 127.0.0.1:5902 0.0.0.0:* LISTEN 3504/qemu-dm-15 tcp 0 0 0.0.0.0:5903 0.0.0.0:* LISTEN 5065/vncterm tcp 0 0 10.31.0.21:59994 10.31.0.24:443 ESTABLISHED 8288/stunnel tcp 0 0 10.31.0.21:5903 10.31.0.238:39476 ESTABLISHED 5075/vncterm tcp 0 0 10.31.0.21:48259 10.31.0.24:443 TIME_WAIT -
It seems to me that the problem is, that qemu-dm-wrapper is listening on 127.0.0.1 instead of 0.0.0.0. Probably qemu-dm-wrapper was changed by some hotfix.
So to sum up again. Using windows templates after installing hotfixes, VNC get's binded to localhost (127.0.01) instead of binding it to every IPv4: 0.0.0.0. Please advise how to get Windows consoles back or make it listen on all IPs. 

Thank you and Regards,
F.
On 5/11/12 11:40 AM, Rajesh Battala wrote:
Hi France,
For which kind of instances you are getting the error message? [Newly created or old vm's ] 
Were you able to view the console for the system VM's?

Thanks
Rajesh Battala

From: France [mailto:[email protected]]
Sent: Monday, November 05, 2012 3:36 PM
To: [email protected]<mailto:[email protected]> 
Cc: Rajesh Battala
Subject: Re: Make vnc bind to 0.0.0.0 insted of localhost on XenServer 6 with CS 3.0.2 

Hi Rajesh,

thank you for your explanation.
What is the proposed solution to get the proxy web console back/working in CloudStack 3.0.2 web GUI?
Currently i'm getting error: " Unable to start console session as connection is refused by the machine you are accessing" 

Regards,
M.
On 5/11/12 10:50 AM, Rajesh Battala wrote:

Hi  France,
We had modified the vnc config on xenserver hosts to listen on localhost instead of 0.0.0.0 because of security reasons.
In your 3.0.2 environment, CS uses https console url provided by the xenserver host to stream the VNC to the Ajax client. 



Thanks

Rajesh Battala





-----Original Message-----

From: France [mailto:[email protected]]

Sent: Monday, November 05, 2012 3:11 PM
To: [email protected]<mailto:[email protected]>
Subject: Make vnc bind to 0.0.0.0 insted of localhost on XenServer 6 with CS 3.0.2 



Hi,
i'm using CloudStack 3.0.2 (as per last official release) and XenServer 6.0.2-53456p (xenenterprise). 



After installing some of these hotfixes:

[root@x4 ~]# xe patch-list | grep XS6

                name-label ( RO): XS602E002

                name-label ( RO): XS602E005

                name-label ( RO): XS602E006

                name-label ( RO): XS602E003

                name-label ( RO): XS602E001

                name-label ( RO): XS602E008

                name-label ( RO): XS602E009

                name-label ( RO): XS602E004

                name-label ( RO): XS602E007
VNC terminal get's binded to 127.0.0.1 or localhost instead of binding to 0.0.0.0, which in turn means, that newly created (restarted?) virtual instances can not be accessed with WEB proxy console, because connection to hipervisor host is refused. 

[root@x4 ~]# netstat -apltn | grep vnc

tcp        0      0 127.0.0.1:9504 0.0.0.0:*

LISTEN      11695/vncterm

tcp        0      0 127.0.0.1:9505 0.0.0.0:*

LISTEN      12269/vncterm

tcp        0      0 127.0.0.1:5901 0.0.0.0:*

LISTEN      8128/vncterm

tcp        0      0 0.0.0.0:5902 0.0.0.0:* LISTEN

28166/vncterm

tcp        0      0 127.0.0.1:5903 0.0.0.0:*

LISTEN      11251/vncterm

tcp        0      0 127.0.0.1:5904 0.0.0.0:*

LISTEN      11695/vncterm

tcp        0      0 127.0.0.1:5905 0.0.0.0:*

LISTEN      12269/vncterm

tcp        0      0 127.0.0.1:9501 0.0.0.0:*

LISTEN      8128/vncterm

tcp        0      0 0.0.0.0:9502 0.0.0.0:* LISTEN

28166/vncterm

tcp        0      0 127.0.0.1:9503 0.0.0.0:*

LISTEN      11251/vncterm



#####

How can i make newly created (restarted?) virtual instances to bind to

0.0.0.0 again?

#####





We've had a lot of other problems after the upgrade(s), most of them

solved bo manually copying files from cloudstack management server to

hipervisors. Like: NFSSR.py to fix storage template problems. This

problem might be related to files:

/opt/xensource/libexec/vncterm-wrapper or qemu-dm-wrapper but i don't

want to screw around with these on production system myself.

Please advise. Thank you.



Regards,

France

Reply via email to