Queueing for clients to be available?

Hi all,

My linkerd setup is discovering the services via consul and linkerd is deployed as a “side-car” model along with the application. Now, I think this is what i observed during the deployment - linkerd service started and setup the listening ports, the application server started successfully and started getting traffic. However, for a second or two, I got this error:

users_get: Internal error processing users_get: 'Failure(connection timed out: /<ip>:11500 at remote address: /<ip>:11500. Remote Info: Not Available, flags=0x09) with RemoteInfo -> Upstream Address: /127.0.0.1:47982, Upstream Client Id: Not Available, Downstream Address: /<ip>:11500, Downstream Client Id: #/io.l5d.consul/.local/<service name>, Trace Id: 732ad639de522b67.732ad639de522b67<:732ad639de522b67 with Service -> 0.0.0.0/11500'

The above is the service I am trying to connect to.

It seems like this is because, linkerd hasn’t yet been able to get a downstream service client to send the request to? Is there a way to say “linkerd, allow connections to be queued for n seconds” before erroring out?

Thanks for any hints.

Best Wishes,
Amit.

Linkerd has a connectTimeout which is the maximum amount of time it will spend establishing a connection. I believe the default value is 1 second but we don’t currently expose a config property for configuration this. We’d happily accept an issue or pull request for this, though.

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.