by GaD » Sat Jul 12, 2014 1:55 pm
Some hypervisors do a poor job at hypervising (??..., jaja, I think I just created a new word..., looks funny...). I've seen this behavior with Hyper-V from Microsoft. I've had the worst experience with it and, to me, its crap! ...sorry Microsoft supporters... I've come to find that the I/O may cause these problems when you are doing call recording. I've also seen lots of lost packages and high latency on ckeint servers (my Vicis, in this case) that are not so ...'populated..... Not so many agents are logged in, but I experience these types of error. I've used VMware, KVM, QEMU, Virtualbox to virtualize Vicis and they all work pretty well, but Hyper-V adds something to the equation that makes it very, VERY nasty to deal with. Hence my question of virtualization.
On other topics..., you might want to make sure that you are syncing time across all servers involved. I usually sync on of them to the world's NTP and the rest I sync to that internal server, make sure that a network hiccup to the Internet won't affect my timing.... Check DB load and ammount of sessions, make sure you're not reaching your limit. Check f your network latency is, at least, descent. You may not be loosing packages, but your latency (time it takes for a ping to 'return') will generate this problem as well.