tcp health check failing to get a response from redis server

[Eric Abbott]

I'm attempting to create a tcp health check to be used in an internal tcp load balancer.

Using "echo PING | nc -q 1 localhost 6379" I can see with tcpflow that the response +PONG

is being sent.

However with the request: PING and response: +PONG set in a tcp health check,

After setting up the firewall rules correctly, I can see the PING show up from the heath

check IPs but I do not see a response being sent by the redis server. I assume that

either a) the connection is being closed by the client before the response is sent or

b) the redis server sees the request as bad and closes the connection.

Perhaps the request needs an actual hard return? Sending PING\n also failed.

Has anyone had success creating a tcp health check against a redis backend?
Any suggestions?

Thanks

[Karthick (Cloud Platform Support)]

Hello Eric,

Could you give some more information about the use case of health check to Redis and how is your health check looks like?  Meanwhile, I would also suggest you to check at stackoverflow for additional insight. 

[Eric Abbott]

Hi Karthick, thanks for the response.

For the use case, I'm setting up an SSL tunnel to RedisLabs. Rather than having a SPOF

I'm attempting to set up a pair of SSL tunnel instances fronted by a load balancer. I'm attempting

to set up the tcp health check to hit the redis instance on the far side of the SSL tunnels rather than

just detect if the SSL tunnel hosts are accepting tcp connections.

The stackoverflow question linked was indeed attempting to match a substring, however

my use case is attempting a full string match. In addition, as mentioned, I'm utilizing tcpflow

on a redis host and am not seeing any output at all being returned to the health check requests.

My assumption is that the health check is sending the --request string PING and waiting, however

the redis server is waiting for perhaps a \n before it will process the incoming string as a request.

Then the health check times out and closes the connection and thus never gets a response to

process.

is there a way to have the tcp checker send a \n (CR) or \r\n (LFCR) in it's --request string?

[Karthick (Cloud Platform Support)]

Hello Eric,

I see that you have opened Feature Request under Public Issue Tracker for including strings (CR+LF) in the request Flag. I believe that that this could be one of the issues that the engineering is working on. However, future discussions will be held there.