Context
Users may encounter issues when running self-hosted instances of LangSmith.
Answer
The first step in troubleshooting is to gather important debugging information about your LangSmith deployment. Service logs, kubernetes events, and resource utilization of containers can help identify the root cause of an issue.
You can run our k8s troubleshooting script which will pull all of the relevant kubernetes information and output it to a folder for investigation.
The script also compresses this folder into a zip file for sharing. Here is an example of how to run this script, assuming your langsmith deployment was brought up in a langsmith namespace:
bash get_k8s_debugging_info.sh --namespace langsmithYou can then inspect the contents of the produced folder for any relevant errors or information. If you would like the LangSmith team to assist in debugging, please share this zip file with the team.
Note that when you run kubectl commands that bring down the original pod like delete or rollout restart, critical logging will be lost when the old pods are replaced. Ensure that prior to replacing the old pods that the script is run on the affected pods.
How to share?
If using Slack of Teams, please reply within the thread
If using email or have restrictions on sending files, please share it to us by creating a ticket at: https://support.langchain.com/
Note: These diagnostic files do not include any trace data, including sensitive information exchanged between your users and models.