Closed Bug 909248 Opened 11 years ago Closed 11 years ago

Logging off with the VNC client on mm-win-81-32-* nodes disconnects the nodes from mozmill-ci

Categories

(Mozilla QA Graveyard :: Infrastructure, defect)

x86
Windows 8
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: whimboo, Assigned: whimboo)

References

Details

As what I have seen right now is that whenever I close my VNC connection to the above mentioned machine, the node disconnects from the mozmill-ci cluster. It could be that we auto-logoff in those cases and close the Jenkins connection. I will investigate.
Interestingly this only happens since I brought up the other Windows 8.1 nodes a couple of minutes ago. Might be a strange side-effect with the other machines.
So for the mm-win-81-32-1 node I can see the following in the Event viewer:

Session "LogonUILog" failed to start with the following error: 0xC0000035

Surprisingly the other two 32bit boxes also started to fail, but not during log off but during log on. So I have checked the machine names and have seen that I accidentally gave the 2nd 32bit machine the same hostname as for the first one. I corrected that and restarted the machine.

With that change the logon/logoff issue still remains. Not sure yet if the above mentioned LogonUILog failure is related here.
Summary: Logging of from VNC client on mm-win-81-3-1 disconnects Jenkins node → Logging off/in with the VNC client on mm-win-81-32-* nodes disconnects the node from mozmill-ci
Looks like that all 32bit Win 8.1 nodes are affected by the logoff issue now.
Summary: Logging off/in with the VNC client on mm-win-81-32-* nodes disconnects the node from mozmill-ci → Logging off with the VNC client on mm-win-81-32-* nodes disconnects the nodes from mozmill-ci
Blocks: 900295
Adrian, it looks like I need your help here. I cannot figure out what's going on or what kind of failure it is triggered by which tool or system application. I removed everything from mm-win-81-32-1 even all additionally installed Windows services. But the problem persists. At latest I accidentally deleted the network device, so now I'm no longer able to connect.  :(

Do you have an idea what's going on here? Google was not much of a help regarding the above mentioned failure case. Thanks.
Flags: needinfo?(afernandez)
NIC re-added to mm-win-81-32-1 and mac address corrected in inventory to reflect the change. VM is now back online.

As for your issue, perhaps best to troubleshoot "live", as right now only seeing mm-win-81-32-3 as being marked as offline in the dashboard (mm-ci-master.qa.scl3).
Flags: needinfo?(afernandez)
Talked with Adrian on IRC and as we figured out the VM was not fully reset. Given that we do not create a snapshot after the creation from the template, we will have to re-create the VM. Can you please do that Adrian? Thanks!
Flags: needinfo?(afernandez)
(In reply to Adrian Fernandez [:Aj] from comment #5)
> As for your issue, perhaps best to troubleshoot "live", as right now only
> seeing mm-win-81-32-3 as being marked as offline in the dashboard
> (mm-ci-master.qa.scl3).

No, this is mm-win-8-32-3 which has another problem and will be covered on another bug as talked on IRC. It has to stay offline.
ok, so as far as this bug is concerned, we are all set here or something pending?
Flags: needinfo?(afernandez)
Adrian, the VM has still to be recreated as we talked about on IRC (see comment 6). And that hasn't been done yet. So please replace mm-win-81-32-1 from the template.
Flags: needinfo?(afernandez)
mm-win-81-32-1.qa.scl3.mozilla.com re-deployed from template and proxy settings confirmed there (no latest updates applied etc)

Previous VM renamed to mm-win-81-32-1.qa.scl3.mozilla.com_909248
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Flags: needinfo?(afernandez)
Resolution: --- → FIXED
err, re-opening as original issue still needs to be checked/fixed.
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
So I tested the latest version of UltraVNC on this box and it works fine. No more disconnect of Jenkins when I login. So I would say we change all of our windows machines to make use of VNC which hasn't been done initially in SCL3. And as Clint said, all of our testing machines use VNC. So we should do the same for the Mozmill machines.

The only question is which VNC software is the right one. Clint or Bob, do you know which one we make use of on the testing machines?
Status: REOPENED → ASSIGNED
Oh, looks like it's indeed UltraVNC:
https://wiki.mozilla.org/ReleaseEngineering/How_To/Loan_a_Slave

So I can go ahead and update all the Win 8.1 machines. If it proves to work, I will file a new bug to get all the Windows machines updated.
All hosts have been updated and are working fine. I will file a follow-up bug to get all the other machines and templates updated.
Status: ASSIGNED → RESOLVED
Closed: 11 years ago11 years ago
Resolution: --- → FIXED
Product: Mozilla QA → Mozilla QA Graveyard
You need to log in before you can comment on or make changes to this bug.