I have kind a question about cross-host communication between containers.
- I have 10 AWS-EC2 Instances with Ubuntu15.10, Docker 1.12.5 and Rancher 1.2.1
- By set up all clusters (kafka, zookeeper and applications using kafka), everything is working fine
- After three days I can observe following behaviour:
** I name the ec2 instances, where kafka-containers are run, kafka-1…kafka-x
** Communication from kafka-1…kafka-x containers to other EC2 instances’ containers is working
** Communication from kafka-1 containers to kafka-2,…,kafka-x is not working
** ping command confirms the issue: I can ping from kafka containers to other instances’ containers. If i want to ping from kafka-1 containers to other kafka-instances containers, it is not working
** I use managed network for kafka containers. Default Network for every stack/service
- A restart of ipsec (rancher/net:v0.7.5) will restart the healthcheck and kafka-container too and the problem is gone. After 3 days it’s happening again.
- I went through http://docs.rancher.com/rancher/v1.2/en/faqs/troubleshooting/ and everything seems ok
- No limitation about security groups and vpn rules on AWS Instances
How can I avoid this behaviour?