Description: Since migrating our Nexus Repository OSS (running in swarm mode) to a new cluster on Red Hat 9 with Docker 26, we have been experiencing significant performance degradation during artifact uploads. Initially, the upload speed is around 26MB/s, but it gradually decreases to 250KB/s over time. Restarting Nexus temporarily restores performance, but it degrades again within one or two days.
Steps to Reproduce:
Deploy Nexus Repository OSS in a Docker Swarm environment on Red Hat 9.
Start uploading artifacts to the repository.
Monitor the upload speed over time.
Expected Behavior: The upload speed should remain consistent and not degrade over time.
Actual Behavior: The upload speed starts at 26MB/s but gradually decreases to 250KB/s. Restarting Nexus temporarily restores the speed, but it degrades again within one or two days.
Workaround: To mitigate the issue temporarily, we implemented an automatic daily restart of Nexus. This solution worked for about a week, but now our reverse proxy returns errors because Nexus exceeds the read timeout, despite our attempts to adjust it.
Feature or Behavior Required: Consistent and reliable upload performance is essential for our CI/CD pipeline and overall development workflow.
Nexus Repository Deployment:
Version: 3.68.1 (upgraded to 3.70.0)
Operating System: Red Hat 9
Docker Version: 26
Database: OrientDB
Additional Information:
Numerous resolutions attempted: adjusting open file limits, updating Nexus, using different base images (Alpine, various Java versions), bypassing reverse proxy and swarm managers, verifying MTU settings, increasing container memory limits (from 10GB to 16GB), adjusting heap and direct memory settings.
Network tests indicate the issue is specific to Nexus. Other services (e.g., GitLab) migrated without issues.
Interesting observation: Downloading a large file using curl from within the Nexus container is extremely slow, but normal from another container on the same swarm node.
Description: Since migrating our Nexus Repository OSS (running in swarm mode) to a new cluster on Red Hat 9 with Docker 26, we have been experiencing significant performance degradation during artifact uploads. Initially, the upload speed is around 26MB/s, but it gradually decreases to 250KB/s over time. Restarting Nexus temporarily restores performance, but it degrades again within one or two days.
Steps to Reproduce:
Actual Behavior: The upload speed starts at 26MB/s but gradually decreases to 250KB/s. Restarting Nexus temporarily restores the speed, but it degrades again within one or two days.
Workaround: To mitigate the issue temporarily, we implemented an automatic daily restart of Nexus. This solution worked for about a week, but now our reverse proxy returns errors because Nexus exceeds the read timeout, despite our attempts to adjust it.
Feature or Behavior Required: Consistent and reliable upload performance is essential for our CI/CD pipeline and overall development workflow.
Nexus Repository Deployment:
Additional Information: