We are experiencing intermittent periods where we will get a considerable number of requests that fail with this error:
“1 error occurred: * gocql: no response received from cassandra within timeout period”
When looking through the documentation for the Cassandra storage backend I see that the connetion_timeout value is available.
There does appear to be an issue where the default is set to 0s that is in PR. We can set that manually and will be trying that. However, I do not think that our issue is related to connect timeout but is rather a request_timeout
Reading through the gocql documentation there appears to be two timeout defaults:
Timeout time.Duration // connection timeout (default: 600ms) ConnectTimeout time.Duration // initial connection timeout, used during initial dial to server (default: 600ms)
Though it seems like if you don’t provide a config that vault literally defaults to a 0s connect timeout.
My questions: Does vault provide overriding a request timeout (similar to cqlsh) or does it only use the one connect timeout? Is the connection_timeout literally our issue and we’re falling victim to the issue linked above?
Any discussion/insight would be helpful. Thanks