Help Center> ModelArts> Troubleshooting> Inference Deployment> Service Deployment> What Can I Do if the Memory Is Insufficient?
Updated on 2024-04-30 GMT+08:00

What Can I Do if the Memory Is Insufficient?

Symptom

  • The deployment or upgrade of a real-time service fails and information similar to the following is displayed in the event.
    Figure 1 Example 1 of a message indicating insufficient memory
  • An alarm is generated for a running service, and the following suggestion is displayed in the event: "Insufficient memory, please increase memory."
    Figure 2 Example 2 of a message indicating insufficient memory

Possible Causes

  • If this message is displayed during deployment or upgrade, the memory size of the chosen compute node is insufficient for the application deployment, and you need to increase the memory.
  • If an alarm is generated for a running service, memory overflow occurs due to code problems, or the service usage is too large so the memory requirement increases.

Solution

  • When deploying or upgrading a real-time service, select a compute node with larger memory.
    Figure 3 Compute node specifications
  • If an alarm is generated for a running service, check whether memory overflow occurs due to code problems, or more memory is required due to heavy service usage. If more memory is required, upgrade the real-time service and select a compute node with larger memory.