SIP Server HA poll

[bublepaw]

Hi,

I was just wondering what mechanism are used by all of us for HA when SIP Server is being deployed

Pawel

[Kubig]

Hi Pawel,
it might be interesting to know the advantages and disadvantages of each HA methods.

[bublepaw]

I will post this one at the end of the poll

[Kubig]

Great! Stay tuned :-)

[smile]

i'm using ip takeover and ms nlb, because it depends on operating system where sip server installed. of course i prefer *nix, that's why ip takeover was selected
for my opinion external nlb (like f5) is better than scripting, but it's a bit expensive. The biggest disadvantage in ip takeover is that after switch-over we hate to announce new mac address to nearest router.

i have another one way, but it's not tested yet - using vrrp daemon. what do you think about it?

ps
btw, what's about resource manager HA?

[Kubig]

And what do you think about Cisco ACE? Does anybody experience with this balancing?

[bublepaw]

Cisco ACE is not supported due to limitation on ACE side. Cisco doesn't understand HA on Genesys side and when primary SIPS fails it will drop exsting transactions and only new transactions will be forwarded to backup node which was promoted to primary. This was result of testing done around 2 year ago. I am not sure if it was changed

[bublepaw]

I am suprised that NLB is leading in our poll. Anyway here is my summary of all three methods:

NLB
  pros:
    widely deployed
    handles total host failure on Windows transparently
  cons:
    complexity
    dedicated vlan
    Windows only
    single subnet

IP takeover
  pros:
    all OS'es supported
    simple
  cons:
    some problems with handling total host failures on Windows
    single subnet

F5
  pros:
  fastest of all methods
  cons:
  complex setup
  additional cost for customers who are not using F5 for soemthing else already

In F5 I wanted add most reliable but recently I met customer who wants to move away from F5 because of stability issues on F5 side.

[Fra]

I'm personally surprised of the high percentage of those using IP takeover..so far I haven't met anyone using it.

Re NLB: I think the high percentage is due to the fact that it was the first supported HA method for Windows. Also, it comes with Windows OS, so SI / customers may have some non-Genesys internal resources that can implement / support it. As an additional con, I'd add 'potential slowness to converge'.

Re F5: I think you've swapped pros & cons I'm surprised about that customer's comments, as F5s are in general well known for reliability.

Fra

[Kubig]

[quote author=Fra link=topic=7451.msg31927#msg31927 date=1349738522]
I'm personally surprised of the high percentage of those using IP takeover..so far I haven't met anyone using it.
[/quote]
We use IP takeover for multi-site and multi-tenant call center with up to 300 agents online and works fine :-)

[bublepaw]

[quote author=Fra link=topic=7451.msg31927#msg31927 date=1349738522]
I'm personally surprised of the high percentage of those using IP takeover..so far I haven't met anyone using it.

Re F5: I think you've swapped pros & cons I'm surprised about that customer's comments, as F5s are in general well known for reliability.
Fra
[/quote]

I was posting after long day - now it is fixed. As for IP takeover I am suprised that it is not used that often as it is much easier to deploy.

I was also suprised by F5 issue but it was only related to SIP part of load balancing - same device was working perfectly fine with web

[naviat]

My solution use IP takeover but the script run for RHEL is not working  fine, when i switchover, application has service unavailable.

Anyone can help me ??

Here is my script:
# Script set_ip_up.sh for up virtual IP
#!/bin/sh
Logpath=/opt/gcti/logs
echo -e "\n--------Bringing VIP Up----------" >> $Logpath/Takeover_On.log
date >> $Logpath/Takeover_On.log
Interface=eth0:1
Virtual_IP=10.84.70.14
#echo $Virtual_IP
# For example Virtual_IP=172.24.133.254
Gateway=10.84.70.1
#echo $Gateway
ping_count=1
loop_count=20
for (( i=1; i<=$loop_count;i++ ))
do
result=$(ping -c $ping_count $Virtual_IP | grep "bytes from")
if [ "$result" ]
then
echo "$i ping reply got from the HA pair" >> $Logpath/Takeover_On.log
sleep 1
else
echo "VIP disabled in the HA pair" >> $Logpath/Takeover_On.log
fi
done
/sbin/ifconfig $Interface $Virtual_IP up
/etc/init.d/network restart
arping -s $Virtual_IP $Gateway -f
echo -e "\n--------VIP enabled in this machine----------" >> $Logpath/Takeover_On.log





==========================

Here is my script down virtual IP

Logpath=/opt/gcti/logs 
echo -e "\n----------Bringing VIP down---------" >> $Logpath/Takeover_Off.log
date >> $Logpath/Takeover_Off.log
Interface=eth0:1
Virtual_IP=10.84.70.14
/sbin/ifconfig $Interface down
sleep 10
ping -c 2 $Virtual_IP
echo -e "---------VIP disabled--------------" >> $Logpath/Takeover_Off.log

[smile]

there is a one funny thing. if you really want to check how good your script, do the following test:
1. unplug network cable from server
2. wait ~ 1 minute while switch-over ends
3. plug network cable back
4. enjoy


this test it not the same when you manually turn off application or kill server power.