[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Xen, Squeeze, and Beyond

Marco d'Itri wrote:
> On Feb 26, Luca Capello <luca@pca.it> wrote:
>>>> 5) Do we recommend that new installations of lenny or of squeeze avoid
>>>> Xen for ease of upgrading to squeeze+1?  If so, what should they use?
>>> It depends. KVM in lenny is buggy and lacks important features. While it
>>> works fine for development and casual use I do not recommend using it in
>>> production for critical tasks.
>> Is the qemu-kvm backport the "correct" solution, then?
> You also need a recent kvm driver in the host, so probably you should
> just use a newer kernel at least in the host.
> I have tens of lenny guests (with their standard kernels) on RHEL 5.4
> hosts and so far I had no issues, but so far most guests are not heavily
> loaded.
The biggest problem for me (and work) right now is live migration. Among
other things, there are known, fixed-but-not-yet-upstream issues with
the KVM paravirt clock which prevents live migration from working.
Unfortunately, there's also no way to disable the paravirt clock from
the host unless you patch qemu or wrap ioctl() and filter the capability
(I did the latter).

Beyond that, I've also seen filesystem corruption when using live
migration and the filesystem cache hasn't been disabled -- an almost
undocumented directive of libvirt's XML.

All in all, I'm wondering how people can call this "stable".


Reply to: