7 servers but running into dead air very often

[jshasteen]

So I built this cluster:

Vicibox 9
VERSION: 2.14-718a
BUILD: 190902-0839

DB:
Xeon Hex Core E5-1650v4 3.6 GHz
24 GB DDR4-2400 ECC
LSI MegaRAID 9361-4i
4 X 240 GB Samsung 883 DCT SSD RAID 10

WEB:
Xeon Hex Core E5-1650v4 3.6 GHz
16 GB DDR4-2400 ECC
240 GB Samsung 883 DCT SSD

DIALERS 1 - 5:
Xeon Quad Core E3-1230v6 3.5 GHz
16 GB DDR4-2400 ECC
240 GB Samsung 883 DCT SSD

I was hoping to be able to support 120 (then 200 after some upgrades) users with an outbound dial ratio of 5 without any issues. Mainly I am running into LOTS of dead air calls. Firewalld is setup properly, manual dial test call after test call works fine. Packet captures look good. What can I do?

Should I add another CPU to the DB? Another CPU to the Web server. Add an additional web server and load balance? Balance agents across a sixth dialer?

I'm pretty confident it's not carrier but of course not 100%. Any feedback would be greatly appreciated.

[garski]

Hi Jhasteen,

You already fix this? We got the same issue. We are using Vicibox 9 on Dialers. It only happens on AMD (8369). We even tried changing carriers but has same issue. 1 DB and 5 Dialing Servers. And no load issue. even 1 -2 agents only experience same thing.

[jshasteen]

What type of latency between your agents and the dialers?

[garski]

I have 160ms - 210ms for the agent side. It only happens whenever we are using AMD 8369. I am using Asterisk 13. Only happens when we upgrade everyone to Asterisk 13 from 1.8

[alo]

What version of asterisk were you using?

We noticed this too, and downgraded to asterisk 11.25.3 but it looks like asterisk 13.27.0 might be better then the other versions of asterisk 13 we were trying before.

[garski]

13.27.1

Are you on v11? How was it compare with 13?

[alo]

11.25.3-vici and the dead air calls completely stopped

[jshasteen]

Interesting. Can Vicibox 9 use that version during the install? I wonder If I can downgrade an existing install.

[mcargile]

I think what you are running into is the major changes that happened between Asterisk 11 and 13. Prior to Asterisk 13, when a call had no audio it could not be routed to the agent's. Instead they would just get statused as LRERR. With Asterisk 13 they do get routed to the agents. This is why you see and increase when switching. Asterisk 13 is not causing you to have more Dead Air calls. It is just letting them through and downgrading will not actually help your problem. Sure the agents will stop getting these calls, but you will still be wasting these leads.

Most likely one of two things is happening. Either your carrier is having MAJOR issues with Dead Air or if you are calling the US your caller id number has been flagged as a SCAM. As soon as you get flagged as a scam the carriers will deprioritize your traffic and start routing you through crappy lines. Some will even flat out purposefully route you to dead air in an effort to waste your time. If you are calling the US, change your caller id number and see if that makes things better. If not create a status like DAIR (Dead Air) and have the agents start statusing calls as it. Use the export calls report to gather those examples and bombard your carrier with them.

[alo]

I was actually starting to wonder today about that. Totally makes sense. Thanks for the explanation!

[jshasteen]

Interesting. Thanks for the detailed explanation Mcargile

[garski]

uisually we are getting 3% - 10% amount of calls on DeadAir. We will try to downgrade it to v11 as per suggestion by alo to see if there is an impact.