Archive for March, 2012

So since euca2ools did not have enough work for me to continue at it for the rest of the semester, I decided I ll simply jump on our main project the Eutester project. It has been a good day so far, most things went well I heard some people had problems with version mismatches so I downloaded and installed everything from source code their most up-to-date versions of “all the things” (boto/m2crypto/paramiko/eutester). Reading through the wiki that the eutester group of guys has set up on cs.worcester.edu, and following  some links to blogg posts from the developers, I managed to run some tests on our newly installed Eucalyptus 3 cloud which is still not able to keep a VM instance running.

>>> eucaops.get_emi()
Image:emi-238B37D9
>>> emi = eucaops.get_emi()
>>> reservations = eucaops.run_instance(emi)
Attempting to run instance-store image Image:emi-238B37D9 in group default
Beginning poll loop for the 1 found in Reservation:r-860340D3
Beginning poll loop for instance Instance:i-0696428B to go to running
Instance(i-0696428B) State(pending), sleeping 10s
Instance(i-0696428B) State(pending), sleeping 10s
Instance(i-0696428B) State(pending), sleeping 10s
Instance(i-0696428B) State(pending), sleeping 10s
Instance(i-0696428B) State(pending), sleeping 10s
Instance(i-0696428B) State(terminated) Poll(5) time elapsed (51)
[CRITICAL]: [TEST_REPORT] FAILED: Instance:i-0696428B did not enter the proper state and was left in terminated

 

After some looking around found out about the existence of nc.log file, and here is the portion of the fail runs:

[EUCAFATAL] Failed to connect to qemu:///system
[i-16F93EC9] could not contact the hypervisor, abandoning the instance
[i-16F93EC9] state change for instance: Staging -> Shutoff (Pending)
doDescribeResource: cores=1/2 mem=1604/1732 disk=11/13 iqn=iqn.1994-05.com.redhat:262aa1ddbe2
[i-16F93EC9] Pending pub=0.0.0.0 priv=0.0.0.0 mac=d0:0d:16:F9:3E:C9 vlan=-24 net=-1 plat=linux vols=
[EUCAFATAL ] Failed to connect to qemu:///system
[i-16F93EC9] cleaning up state for instance
[i-16F93EC9] stopping the network (vlan=-24)
[EUCAWARN  ] vnetStopNetworkManaged(): supplied vlan '-24' is out of range (0 - 4096), nothing to do
[i-16F93EC9] state change for instance: Shutoff -> Teardown (Teardown)
[EUCAFATAL ] Failed to connect to qemu:///system

After trying to work my way through an old bug report and fix on the above problem, I decided I should check if Hardware Virtualization, which turned out it was the right idea, because in this case it is turned off in the BIOS.

Advertisements

So some time during week 4 of the project I had some down time, and I decided to learn how euca2ools worked. Simple command-line tools yet still very powerful as well. However recently we run to some problems with the version mismatches; so here it goes: “how to install everything from source with the latest versions”:

Very easy to install from source:
NOTE: make sure to remove older versions of the software.

Download newest boto: https://github.com/downloads/boto/boto/boto-2.3.0.tar.gz

cd /download/location
tar zxvf boto-2.3.0n.tar.gz
cd boto-2.3.0
sudo python setup.py install

Download newest M2Crypto: http://pypi.python.org/packages/source/M/M2Crypto/M2Crypto-0.21.1.tar.gz

cd /download/location 
tar zxvf M2Crypto-0.21.1.tar.gz 
cd M2Crypto-0.21.1 
sudo python setup.py install

Download euca2ools from: https://github.com/eucalyptus/euca2ools

cd /download/location
cd euca2ools
sudo make

And that should be it.

It is finally getting there but still with struggles, right now for example while I am writing this I wish I could instead be connecting to our own cluster and uploading images but port 8773 still seams to be closed when trying to look in from the outside, so I can’t connect to try and work on an instance. While earlier today Long was able to upload kernel images to the server it looks like there is a problem when trying to actually start up an instance on the server which most likely proves right our fears of hardware support for hypervisor being turned off on the BIOS. One might think “big deal” just turn it on; however these are stoke dell machines we are using the BIOS options are basic at best, so we will most-likely have to install things from scratch trying to install an other BIOS on the mother board and finally unlock the hardware support.

And finally we have our definite proof that kvm support is disabled on the BIOS, and here goes an other week without having a working cloud.