some environments are on slow connections or sometimes have to switch over to a slow or wireless connections (3G/4G) which requires a custom cam.cfg change for reliability, but what is the best way to account for that?
Any supported CA Client Automation environment
The default protocol used by ITCA is UDP because it is fast and light, however it is also quite unstable in some environments, particularly those with slow connections. Configuration changes to CAM's default settings can sometimes stabilize CAM and allow UDP to continue to be used and benefit from it's speed.
There are several general considerations when deciding at what level to set a configuration value like 'Fragment_Size=1024' in a custom CAM.CFG file (The choices being agent, SS and/or DM level)
1. How many agents are potentially impacted by this problem
2. How heavily loaded are the SS machines
3. How often are very large packages pushed to these machines
4. How much time is typically acceptable for packages to be delivered
In this scenario we will assume the following details in relation to the above:
1. Many to most agents are impacted
2. A relatively small number of agents are pointed to the DM directly (<1000)
2. Several SS machines are heavily loaded (~1000+) while many others have small numbers of agents (<100) reporting to them (and may be remote systems)
3. Packages are typically small (~200mb or less) with larger packages going out only infrequently.
4. Most packages have ample time to distribute (5-8 hours typically)
In this case, the recommendation would be the following:
1. Place a custom CAM.CFG on all Scalability Servers rather than on agents
2. Convert the High-Volume Scalability Servers to using CAM over TCP for communication with the DM ONLY and UDP to all other systems.
3. Place a custom CAM.CFG on the Domain Manager to set the UDP fragment size, so agents and remote SS machines will connect via that method.
1. Place the custom CAM.CFG file on all Scalability Server systems and the Domain Manager
2. For the High-Volume Scalability Servers, add a line like the following to the *PATHS section of CAM.CFG to force those machines to communicate with the Domain Manager using TCP, but still be able to the fragment_size=1024 setting and UDP for all other end points:
<DM NAME or IP> PROTOCOL = TCP PORT = 4105
3. Ensure that TCP port 7163 is open between these systems and the DM as CAM 4105 TCP traffic is intercepted and handled by the PMUX 'SHIM' dll and broadcast to the remote systems Port Multiplexor over that port before being converted locally back to 4105 and sent to CAM on the target machine.
4. All other Scalability Servers are remote and lightly loaded so there should be only minimal impact of them using UDP for communication with the DM since UDP is faster and lighter which should compensate for the smaller fragment size vs TCP which is a heavier protocol that will not have an overall benefit for those systems if UDP is working.
Finally, it is important to know that these changes will not take effect until CAF and CAM are fully recycled or the system rebooted, so keep that in mind when deploying these changes.