JFrog ML Troubleshooting

Symptom: Increase in 429 errors

A high error rate due to429 error(too many requests) is caused by a shortage of resources.

Solution

This issue can be solved by either scaling vertically or horizontally.

• Vertical scaling

  To scale up, try adding more compute power to your current machines.

  • Add more resources (CPU / RAM).
  • If CPU / memory aren't fully utilized, increase the number of concurrent workers.

• Horizontal scaling

  To scale out, try adding additional nodes or machines to your infrastructure to cope with the new demand.

  • Manually add more pods (increase from 4 to a higher number).
  • Use Autoscaling based on CPU / Memory / Latency.

Symptom: Can't install FrogML SDK

Python SDK deployment on M1.

Solution

Make sure you are not running with rosetta.

Issue: Need to verify CLI and SDK installation.

Solution

Run the following commands to ensure your environment is configured correctly:

• For the CLI: Run frogml --version. If the command is not found, ensure the FrogML CLI is in your system's PATH.
• For the SDK: Run pip show frogml. This confirms the FrogML Python SDK package is installed in your current Python environment.

Symptom: Getting grpc errors (have 'x86_64', need 'arm64')

Solution

When usingcondaand running on Mac M1 CPU, simply run the following command:
conda install -c conda-forge grpcio

If the issue is:dependency_injector/providers.cpython-39-darwin.so' (mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64'))

Do the following:
pip uninstall dependency\_injector
ARCHFLAGS="-arch arm64" pip install dependency\_injector --compile --no-cache-dir

Symptom: Your VPN/Proxy encrypts traffic with additional certificates.

The certificates should be added to the CA certificate file. Typical certificate issues are appearing when configuring your JFrog ML CLI:

Caused by SSLError(SSLCertVerificationError(1, 
'[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)')

Solution

1. Add the CA Certificate to certifi 's Bundle
   JFrog ML SDK relies on Python's certifi library for server certificate validation. To identify the location of your certificate validation file, run the following Python snippet:
   

import certifi
print(certifi.where()

This will output the path to your certificate validation file. Next, append your VPN's or Proxy's CA certificate to this cacert.pem file. Open the file in a text editor with administrative privileges and add the certificate at the end.

2. Verifying the new Certificate CA with OpenSSL

After adding your custom certificate, you can validate the connection using OpenSSL with the following command:

openssl s_client -connect dev-qwak.us.auth0.com:443 -CAfile /path/to/your/cacert.pem

Symptom:

When your cloud resources are behind a private network or VPC, you may need to whitelist specific IP addresses to allow the JFrog ML platform to access them. For example when connecting to a new BigQuery or S3 based Data Source you might get a timeout error.

• You receive timeout errors when JFrog ML tries to access your cloud resources.
• Your cloud resource logs show unauthorized or blocked access attempts from the JFrog ML IP addresses.
• Data transfers or API calls between JFrog ML and your cloud resources are failing without a clear error message.

Solution

You may get these symptoms if your cloud resources are behind a private network or VPC, you may need to whitelist specific IP addresses to allow the JFrog ML platform to access them. For example when connecting to a new BigQuery or S3 based Data Source you might get a timeout error.

To ensure seamless connectivity, please add the following range of JFrog ML IP addresses to your network's whitelist:

`23.21.54.216  

44.212.137.42`