0

We run a TiDB cluster consisting of twelve nodes. The TiDB version is 6.5.0. We have recently added a TiFlash node in order to enable HTAP/analytics processing. Replicating a table appears to work fine. The information_schema.tiflash_replica table shows the respective table as "available". However, when I try to run any SELECT query on the replicated table from TiDB, the MySQL client hangs for a while and then prints the following message:

rpc error: code = Unavailable desc = connection error: desc = "transport: Error while dialing dial tcp 172.18.78.60:3930: connect: connection refused

I have no idea why this happens. Each node (TiDB/PD/KV/TiFlash) is installed on a separate bare metal server in the same network without any gateways or firewalls between them.

Expected result: SELECT COUNT(*) FROM table1 WHERE field = 'value' runs faster when replicated.

junglerider
  • 1

1 Answers1

0

172.18.78.60:3930 is the service held by your TiFlash instance. I think you can first make sure that 172.18.78.60:3930 is able to be connected by the TiDB node.

Login to the machine that your TiDB node hosted, and issue the command:

telnet 172.18.78.60 3930

If your network is good then you should find messages like below.

Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
billmay
  • 1
  • 1
  • Thank you. It was indeed a network configuration problem. For some (obscure) reason, port 3930 could only be reached from localhost, whereas the other TiFlash ports were open. It worked after the cluster topology file was updated with the correct host name. – junglerider Mar 23 '23 at 10:33