Hi Lionel ROCHA,
Basically with chilli would be nearly everything possible. However I do not have the required chili language experience to help but there is a full chilli reference available for download (BCM V12 doc PDFs - BMC Client Management 12.0 - BMC Documentation ). I am pretty sure that there was a chilli script somewhere in the "BCM World" which does the same as the "DHCP option" for the agent configuration / communication. This was used before the "DHCP option" was introduced. Since the "DHCP option" is available since a long time I am not sure that this script does still exist.
Did I understand it correctly that all your clients are connecting to one DNS record like YOUR_Relay.company.com? And this request will be forwarded to your DMZ -> local relay? Wouldn't it be easier to simple use a DNS round robin mechanism? In my opinion this would be more easier and you could easily add more relays if needed without deploying a new configuration or create any ORs / DB triggers and so on.
I will look for chilli.
All clients (4500) are in shop around the world, connected by their own internet connection (no vpn, only ADSL). So I dont have control of their ip.
Then I have one relay on dmz (local).
With this model, I encounter a lot of loss connection with agents. Communication is no stable (tunnel doesn't hold). => is the result on our network expert team.
In my opinion, there are to many clients on the only relay, but the problem is, how a can say on agents to switch between two relay depending on the load (I can't define zones with different agent configurations)
I will search about "DNS round robin", because I don't know.
Rule of thumb for DMZ Relays that must maintain active tunnels due to being behind a NAT address.
Linux Relays = 1800
Windows Relays = 1200
Typically any relay where active tunnels are not required you should have no more than 2500-3000 nodes assigned to that relay for peak performance. These would be for typical WAN connected systems where active tunnels are not required.
Linux devices used as Relays can handle more concurrent connections when properly tweaked (BMC provided scripts to run against the OS) than a Windows device (nature of the beast).
I would recommend assigning a specific Relay using STATIC and then setting up a BACKUP as a failover by creating specific device groups that will carve up your 4500 devices into three distinct groups, each assigned to a specific relay.ini configuration file.
You will need 4 relays with unique external DNS pointers to each of these relays.
If it we me, I would use existing DMZ relay as your failover then
- Set up ADDITIONAL DEVICE ATTRIBUTES to group the three (3) new relays based on your demographics or some other grouping method
- Use a CSV to update these values
- Create three (3) new Op Rule modifying the Relay.ini file to change the MECHINISM entries as mentioned above
- Assign these new op rules to the new device groups based on the values you set in the ADDITIONAL DEVICE ATTRIBUTES.
Always TEST on small sets of devices before pulling the trigger on all devices to ensure that the new rule works as planned.
Continue to use the same rollout that points to existing DMZ Relay and make the change to point to new relays via Op Rules based on the value you set in the Additional Device Attributes
Always add "Restart Agent" as second Step so device will begin using new relay as parent ASAP.
You may wish to work with your L1 BCM Support team, Partner or BMC, before using this solution to confirm I am correct for YOUR implementation. Things also change for each version of BCM and this was a solution used for 12.0-12.1. Maybe BMC no longer believes my number are correct on the number of nodes per relay for a DMZ (Starbucks Mode).
Let me know if you have any questions about this proposed solution to ensure you will always have the ability to stay connected to your end points.
Thank a lot for your answer,
But i don't understand how I can set an ADDITIONAL DEVICE ATTRIBUTES on all devices.
Today, we have one agent deployment with the same STATIC ip (relay). So it's complexe to check all devices and set the "relay attribute".
I have created different group with sql request, who assign rules (relay setup).
The only default is when we have the remote control, and the device is in the process of changing relay, we loss the remote control.
But otherwise it works well.
Please review this document I created for you!
Then just an Op Rule assigned to each subset setting new STATIC Relay. This will then tell the agent which relay it should use!
Thank you very much !