Improve hybrid cloud drivers
|Assignee:||Tino Vázquez||% Done:|
|Category:||Drivers - VM|
|Target version:||Release 4.4|
There are several areas to improve:
- Better and more scalable monitoring for EC2 VMs (change strategy to mimic that of KVM, XEN, VMware)
- Evaluate direct use of API instead of API tools
feature #2289: monitor all VMs in EC2 IM
Get information from all EC2 instances in IM driver and provide
this information to OpenNebula so a VM poll is not executed per VM.
The monitoring will be much faster as VM is very expensive and the time
it takes to get info about one VM is very similar to getting information
about all VMs.
To keep track of the OpenNebula ID in EC2 machines a new tag is added
to EC2 instances (ONE_ID). This is used when monitoring VMs. VMs without
this tag wont be correctly reported to OpenNebula and will be monitored
with the old method (VM poll).
This version is kind of a PoC. The driver should be refactored so it
is similar to other drivers, that is, one script per action. The library
that holds the functionality will be held in a library file in VMM
driver as VMware drivers. Special care should be taken read
EC2 parameters as now is needed by both VMM and IM.
feature #2289: Update ec2 driver to use the ruby sdk instead of the java cli
Capacity and endpoint for the host are now defined in hybrid_ec2.conf based on the host name
feature #2289: Rename ec2.conf (ec2_driver.default) and hybrid_ec2.conf (ec2_driver.conf)
feature #2289: Use deploy id to save and restore an ec2 instance, instead of creating a checkpoint file