Hello! Let's clear up those points regarding Dynamic Agents 🙂
As we know, the dynamic agent is located at the master and also in the FTA right?
Yes, the dynamic agent is located at the master domain manager and fault-tolerant agent.
Purposely dynamic agent for what ?
The dynamic agent handles what the fault-tolerant agent cannot manage:
-
Advanced job types: It is required to run plugins. Fault-tolerant agents mainly run OS-level executable scripts.
-
Resource-based scheduling: It can report its hardware/software characteristics to the dynamic workload broker, allowing jobs to be sent to any agent that meets specific requirements.
-
Simplified communication: It communicates via HTTP/HTTPS (port 31116 by default), which is more firewall-friendly than the proprietary protocol used by FTAs (port 31111).
-
Automatic updates: Like the FTA, the Dynamic Agent supports centralized upgrades from the Master. However, it typically offers more functionality and receives more frequent maintenance updates. A standout feature is the automatic update of certificates, which significantly reduces manual security overhead.
-
Automatic Registration: Unlike traditional agents that often require manual definition in the database, Dynamic Agents have the capability to automatically register themselves with the environment once the service starts.
Should we startup the dynamic agent in Master and in FTA? (by executed command StartUpLwa)
Yes, you should start up the dynamic agent in master domain manager and fault-tolerant agent by running StartUpLwa.
If we not startup dynamic agent either in Master nor FTA, what could it happen or any differences?
If you skip running StartUpLwa on your master domain manager or fault-tolerant agents, your system's capabilities will be limited:
-
Traditional jobs will still work: Your standard OS scripts and executables will continue to run perfectly fine, provided that the jobs are scheduled specifically on FTA workstations. In this scenario, the core fault-tolerant processes (netman, batchman, and jobman) handle the execution independently.
-
Advanced jobs will fail or hang: Any advanced job type (like a REST or DB job) scheduled on a node where the dynamic agent is down will either fail to launch or remain stuck in a waiting/executing state.
-
Loss of dynamic pooling: The dynamic workload broker will see the agent as "Offline". It won't be able to route dynamic jobs to that specific node, reducing the capacity of your dynamic pools.
You can also go through this interesting blog about Dynamic Agents: Dynamic Agents, what a surprise!
Regards,
Veronica Tesei
------------------------------
Veronica Tesei
------------------------------