Re: [bitfolk] ping monitoring showing packet loss to bitfolk VM

Hi Joseph, On Wed, Oct 26, 2016 at 08:27:08PM +0100, Joseph Heenan wrote: Certainly it seems like there is some delay or loss occurring only between the physical host and the guest, not between the physical host and the rest of the world. Although when running long-term pings once per second I do not myself see actual loss - is the Firebrick considering a high enough latency as a loss? I have seen this sort of thing in the past and it seemed to be related to CPU exhaustion. I had hoped we had left that behind with the most recent hardware refresh, as the per-thread CPU performance was increased quite considerably. The host that your VPS is on averages at about 530% CPU (8 cores, so in theory could go to 800%; and in fact 16 threads so again theoretically even more). It may not be just CPU time that is the issue but something to do with context switches. By default Xen lets whichever guest that gets a real CPU to have 30ms exclusive use of it, so you can imagine that if there's only 8 real CPU cores and you have to wait 5 times before you get scheduled then that will be 150ms before you can do anything, such as respond to a ping. I have some ideas I'd like to pursue on this as much of it is configurable, so I will open a support ticket with you Joseph to discuss further. Also if your friend is interested could you get them to email support(a)bitfolk.com and I'll work through things with them too. Cheers, Andy -- https://bitfolk.com/ -- No-nonsense VPS hosting