Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations Mike Lewis on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

AVAYA Duplex Reliability questions

Status
Not open for further replies.

AlexSki

Systems Engineer
May 6, 2019
17
RU
Hi,
i have two 8730 avaya media servers which working in duplex reliability .There are failover link between servers and links from each MS to switch and one IPSI in G.650.Switch only one. Prepose patch Active MS --- switch is down but failover link is good . Will the active MS remain active -? how the server understands that it is necessary to become active -?
How to properly simulate network and servers failure ? I want to be sure that the duplex reliability works correctly.
 
Duplex CM has a duplication manager service and an arbitration service running on both machines. I believe you can grep -r "ARB" /var/log/ecs/ just to get a look at what the logs are saying. that, or DUPMGR too.

How do you test it? Well, if you have port networks, then at least 1 IPSI cannot be 'ignored in server arbitration'. disp ips 1a, I think it's on the second page. By default, all IPSIs have that at 'yes' which means if any IPSI with that parameter enabled goes offline, your duplex pair starts flip-flopping on the basis that it thinks the other server might have a better chance reaching the IPSI. At least 1 IPSI in the system must be setup for this. So, pulling the NIC on it would make your duplex go into action.

Otherwise, if you type 'server' in the cli and the duplex pair is healthy, happy, refreshed, shadowed, etc, you can trust it to behave within normal parameters. It's changed over time from hardware to software to requiring <8ms to being tolerant up to 60ms. Depending how busy a real system is when a failure occurs, you might see different behaviour the one time it happens when it's busy during the day vs the maintenance you do at night when it's quiet.

ARB is the arbiter service and decides for the server when it should be live or not. Over the years, lots has changed regarding duplication but it's always the same principal - how to best cover the scenario of hardware or network failure. That's pretty much the principal for all failover design - mitigating hardware or network failure.

In any case, duplex CM never covered you from a software bug that made CM crash because you were only ever running a single instance of CM.

There were 3 different kinds of duplication boards over the years for hardware duplication - DAJ1 DAL1 and DAL2. They came with he media servers at different periods in time. Those boards job in life was to keep 1 instance of shared memory - like a shared brain - between the two servers. So, duplex CM covers you if one of the servers died or if the customer LAN connected to 1 server became unavailable. Avaya moved to software duplication probably close to 10 years ago.

On the S8720s and S8730s it was an option to do duplication over ethernet ports once CM 5 came out around 2009. I believe duplex 8720s would have required hardware duplication before CM4 or 5 but could be enabled for software duplication once R5 came out. That means you didn't need to spend retail $5000 per board per server.

Once release 6 came out, Avaya was moving to 'common servers'. They called the IBM x3550 the S8800. 1U instead of 2U and no special hardware. Take a large call center environment with Voice Portal, Modular Messaging, CM, and a variety of other applications. If you were the field service manager at Avaya, you had a real hard time keeping a list of every power supply or stick of ram or hard drive you had to get delivered in under 4 hours to be able to keep such a diverse variety of hardware online. The only the servers had in common was they said Avaya. So while VMware was starting to get popular, Avaya built System Platform on the Xen hypervisor and tried to use that as a common platform on the way to VMware. And now it's on VMware, HyperV, KVM, etc. Today, if you want to span your VLAN across data centers, you can have 1 CM of a duplex in each data center if you keep latency <30ms if you have AES and <60ms if you don't use AES.



Here's some backgound on the DAL boards:

Some background on software duplication requirements:


Background on VMware for CM:

Server separation CM 7.1+ (pg 18)
 
software duplication was available s8720 and s8730 beginning with cm3 and higher around 2005.



A great teacher, does not provide answers, but methods to teach others "How and where to find the answers"

bsh

45 years Bell, AT&T, Lucent, Avaya
Tier 3 for 35 years and counting
[URL unfurl="true"]http://bshtele.com[/url]
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top