-
Bug
-
Resolution: Unresolved
-
Major
-
None
-
Jenkins 1.492 on Ubunta
Jnlp slave on win7 x86
Slave is started in Nimbula cloud - our own plugin is used to start virtual machine. Plugin is not published yet.
But I really don't think trouble is in it.
Steps are:
1. Job is started
2. Our plugin is used to start VM to run job
3. VM starts
4. It connects on start to Jnlp url
5. Build process starts
And.. sometimes ~1 of 10 starts (rather rare) build process hangs.
Console is in such state:
Started by timer
Building remotely on Nimbula-slave-1110 in workspace /jenkins-slave/workspace/PRODUCTION - Sanity
[Throbber non stop here]
Node system info is in attachment.
I would be very much obliged if you can help to understand possible reasons!
[JENKINS-16078] All types of Windows slaves hangs on start in ~1 of 10 build starts
Hi the same happens with other windows slave types:
- Windows Slave started via DCOM (no longer used because horrible to configure, holes in firewall, all bad)
- Windows Slave started via JNLP (no longer used because slave.jar is not updated automatically)
- Windows Slave started via SSH (our current solution: a custom script by us that SCPs slave.jar from the WAR file to the slave and then ssh to the machine and start the slave.jar - we have to do this because Jenkins SSH slaves plugin cannot handle the native Windows cmd32 shell; we don't use cygwin, instead the BitviseSSH server).
We see hangs every few builds, in most cases when the slave already started shortly before. So it seems to be an issue for the first build that starts after the slave started (we use the VirtualBOX VM to wake up a Windows VM from standby, update/start the slave.jar, start the jobs sticked to this VM and after that halt slave.jar and let the VirtualBOX VM go to standby/save state).
Once it happens again, I will request a jstack trace from a second SSH console to the Windows machine.
The Windows machine's config is Windows 7 Professional 64bit with the above BitviseSSH config.
Uwe Schindler
added a comment - - edited Hi the same happens with other windows slave types:
Windows Slave started via DCOM (no longer used because horrible to configure, holes in firewall, all bad)
Windows Slave started via JNLP (no longer used because slave.jar is not updated automatically)
Windows Slave started via SSH (our current solution: a custom script by us that SCPs slave.jar from the WAR file to the slave and then ssh to the machine and start the slave.jar - we have to do this because Jenkins SSH slaves plugin cannot handle the native Windows cmd32 shell; we don't use cygwin, instead the BitviseSSH server).
We see hangs every few builds, in most cases when the slave already started shortly before. So it seems to be an issue for the first build that starts after the slave started (we use the VirtualBOX VM to wake up a Windows VM from standby, update/start the slave.jar, start the jobs sticked to this VM and after that halt slave.jar and let the VirtualBOX VM go to standby/save state).
Once it happens again, I will request a jstack trace from a second SSH console to the Windows machine.
The Windows machine's config is Windows 7 Professional 64bit with the above BitviseSSH config. 
Hi
Still reproducing this issue sometimes. Even more rare than 1/10
Maybe there is way to turn on additional debug info into log? To see on which step exactly hanging happens.
The problem is it happens really rare. And I guess it depends a lot from the time conditions.