Hi guys, we ran 20 meetings with 10-15 people each across two of our autoscaling clusters on aws for approximately 3.5 hours and then saw dropout on our JVB network traffic which resulted in dropped callers/calls. HAproxy and overall network throughput remained consistent which leads me to believe this is not client side. Any thoughts as to what would cause these drops? See screen shots.
What are the resources of JVB and JMS?
We are running c5.xlarge for JMS and c5.9xlarge for JVBs. Total users were 169, with 15-20 per conference (only 10 on video) and we had 4 JVBs running across two availability zones.
Do you have scale-in policy?
IIRC there is a bug which drops all connections when a JVB stopped. Probably it’s on the old stable
Thanks, but I don’t think that created the problem- none of the timestamps for the autoscale out events line up with the drop on the JVB network.
One other thing that I saw on our cloud watch is that the network drop occurred in two separate availability zones at the exact time. Each zone showed the same JVB network drop. Any thoughts?
Are the participants joining from different locations spread in different Internet providers?
@damencho Yes, they are joining from different locations and using different ISPs.