[IGNITE-5873] Zookeeper discovery does not work correctly when isolated clusters got deployed on similar hosts

• Type: Bug
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 1.9
• Component/s: zookeeper
• Assignee: Unassigned
• Reporter: MIKHAIL LIKHASHVA
• Created: 28/Jul/17 17:44
• Updated: 28/Jul/17 17:44

Description

Steps:
1. Start Cluster X on machine A and B (under zookeeper path /xxx )
2. Look at what is registered in zookeeper under /xxx
/xxx/88x8x8 HOST_A:47500
/xxx/88x8x8 127.0.0.1:47500
/xxx/54d32d HOST_B:47500
/xxx/54d32d 127.0.0.1:47500

3 . Start Cluster Y on machine A and B (under zookeeper path /yyy )
4. Look at what is registered in zookeeper under /yyy
/yyy/34564rt HOST_A:47501
/yyy/34564rt 127.0.0.1:47501
/yyy/sdf3232 HOST_B:47501
/yyy/sdf3232 127.0.0.1:47501

Look correct. Shutdown and try to execute steps 1&3 simultaneously, several times, you could end up with situation like this:

/xxx/88x8x8 HOST_A:47500
/xxx/88x8x8 127.0.0.1:47500
/xxx/54d32d HOST_B:47501
/xxx/54d32d 127.0.0.1:47501

/yyy/34564rt HOST_A:47501
/yyy/34564rt 127.0.0.1:47501
/yyy/sdf3232 HOST_B:47500
/yyy/sdf3232 127.0.0.1:47500

So now node could incorrect connect to different cluster. Possible solution to filter out 127.0.0.1 and 0.0.0.0 during start up in registerAddresses of TcpDiscoveryZookeeperIpFinder