Milvus Abort on Startup #36896
Replies: 4 comments 5 replies
-
Does the CPU instruction meet this requirement? Required:
|
Beta Was this translation helpful? Give feedback.
-
Thanks for raising the issue. This is a duplicate of #33964. I've asked around the Milvus maintainers and unfortunately there isn't a support of apptainer a now. Milvus only supports docker container and k8s. We will see if the community has any expert on HPC and Apptainer to help build that feature. Meanwhile I'd suggest to try docker or k8s deployment env. If performance isn't a concern (say the data set is less than a million vectors), you can try https://milvus.io/docs/milvus_lite.md which only requires python env. |
Beta Was this translation helpful? Give feedback.
-
@zwd1208 |
Beta Was this translation helpful? Give feedback.
-
Hi there and thank you for your patience. I'm not very familiar with Apptainer. After studying it carefully, I tried to troubleshoot the issue you encountered in the following environment, but I couldn't reproduce the problem. Below are the operating environment and steps. Could you please provide the specific logs, environment version, and operation process on your side? Version of Apptainer
OS Version
Steps of Operation
2.Build images PS.Need to execute the startup command in the relative path |
Beta Was this translation helpful? Give feedback.
-
We want to run Milvus on our university’s (University of Florida) supercomputer. Unfortunately, the Research Computing department here does not use Docker but Apptainer. We therefore pulled the latest Milvus image from Docker Hub into Apptainer and have been successful starting the milvus-etcd and milvus-minio processes. However, every time we try to start milvus-standalone, the process aborts immediately after it executes the line for milvus-standalone. No error messages are thrown. We have no idea what’s going on and nothing is listed in the core.milvus file. Our process is running with 10g of memory and 1 CPU. We can increase those if needed and ask for more memory, CPUs, and GPUs in our SLURM request.
If anyone has any ideas, we’d appreciate hearing them as we’re not making much progress on this one.
Beta Was this translation helpful? Give feedback.
All reactions