Patching the cloud – Azure failure
Hoff posted some nice comments on the Azure’s failure regarding patching the infrastructure used by cloud services. An interesting conclusion about it is that future patching mechanisms will have to be integrated to VMotion-like features, in a way that when you apply an OS patch to the infrastructure it can dynamically deal with that without disrupting the service. It would be something like this:
- Move the virtualized hosts from one server to the others
- Patch it the “idle” server
- Check if it comes back properly
- Gradually puts back the load on that server and checks if there is any impact from the patch
- If everything is ok, go back to step #1 for the next server – repeat until all servers are patched
I wonder if the guys from Microsoft Update are talking with the Azure team - big challenge for team integration ahead, and business opportunity for patch management companies.
[...] original here: Patching the cloud – Azure failure | Security Balance patch [...]
March 20, 2009 @ 11:08 am
Augusto,
Do not forget to add #6 – REGRESSION TEST. Ideally, any change like that should include explicit regression tests.
March 21, 2009 @ 7:09 am
You know that VMware based clouds will already have the patching thing down since they include VMware Update Manager. It does exactly what you say: move VMs off, patch host, move VMs back on. No disruption to service. Entirely automated. Been in existence and used by hundreds of thousands of customers for a couple of years already. Given that you’ll have a lot more choice using one of the over 100 VMware powered clouds that already has this versus being stuck in 2 MS datacenters without this I think the choice should be clear on where to go.
March 28, 2009 @ 8:26 am