Why Does My Node Not Connect to Spread?

Posted June 14, 2018 by Soniya Shah, Information Developer

If your node is not connected to spread, follow the below steps to troubleshoot the connection issue.
Step Task Results
1 Check whether the spread.conf file in the catalog folder is identical across all the nodes in the cluster. $ cat spread.conf If the spread.conf flies are identical across all nodes, go to Step 2. If the spread.conf flies are not identical across all nodes, do the following:
  • Identify the correct spread.conf file. The file has all the cluster node IP addresses.
  • Copy the correct spread.conf file onto all the node using scp.
2 Check whether the spread port (default is 4803) is listening using either of the following: $ netstat -ap | grep 4803 OR $ telnet <spread IP address> 4803 Each node in your cluster has its own spread port. To verify the port used by spread in your cluster:
  • Look in /database_catalog_directory/spread.conf
  • Look for the Spread_Segment line after the IP mask, such as Spread_Segment
If the spread is not listening, go to Step 3. If the spread is listening, go to Step 4.
3 Check that the firewall is not enabled. # iptables –L If the firewall is enabled, either disable the firewall OR ask the network administrator to open the port.
4 Check whether the /etc/hosts file has all the IP addresses of the cluster. If the file is correct with all the IP addresses of the cluster, go to Step 5. If the file is not correct, fix the /etc/hosts files to include all the nodes in the cluster and go to Step 5.
5 Check whether the UDP transmit packet contains errors. $ netstat -s If the UDP transmit packet contains errors, go to Step 6. If the UDP transmit packet does not contains errors, go to Step 7.
6 Verify the packet contains errors using the following command to look at the UDP transmits. $ /opt/vertica/bin/vnetperf If there are errors contact your local administrator or system administrator.
7 Check whether spread groups form using spuser.
    a. Run the following on all nodes in the cluster:
$ /opt/vertica/spread/bin/spuser -s 4803@host_IP
    b. Execute the following:
$ j test-spread
    c. Verify that the number of members found in the previous step match the number of nodes.
If the number of nodes match, your spread must be connected and the checklist is complete. If the number of the nodes do not match, start this checklist again to find an alternative path. If the number of members continues to not match the number of nodes, contact Vertica support.

Learn More

Learn more about Spread in the Vertica Documentation and Spread Debugging in the Vertica Knowledge Base.