Re: network buffering in fault tolerance

[Jason Wang <jasowang@redhat.com>]

On 2020/9/30 下午1:12, Shivam Mehra wrote:
>
> I came across this documentation with source code for providing 
> network buffering to applications 
> https://www.nfradead.org/~tgr/libnl/doc/api/route_2qdisc_2plug_8c_source.html 
> <https://www.infradead.org/~tgr/libnl/doc/api/route_2qdisc_2plug_8c_source.html>. 
> This network-buffering helps output-commit problem when providing 
> fault tolerance to virtual machines. The output is buffered until an 
> acknowledgement arrives from the backup VM and then released to the 
> external world. So that backup and primary VMs seem consistent 
> externally. Initially developed for XEN VMM to provide fault tolerance 
> to VMs and I think it's now available for |QEMU| too.
>
> Where does the script reside which does network-buffering for 
> checkpoints? and what are the commands to make this happen?
>
> I want to do this network-buffering for packets originating from an 
> application. Is it possible to do it in the same way as above? Does it 
> do any damage to the host kernel? Can I get  a simple working example 
> for this?
>

Hi Shivam:

I haven't gone through the qdisc you mentioned, but it depends on a 
specific host network setup to work if I understand correctly.

Qemu support a native HA (COLO) which has a buffer filter.

Cc Chen for more information.

Thanks