I am not aware of a bug similar to the behaviour that you have just described.
I do have a few questions, though:
What version of Couchbase Server are you running?
What were the number of connections before running the test script, perhaps you could share said test script?
Where exactly are you seeing these connections? Is this the total connections graph in the UI?
Could you run a netstat -anp on the node when you see this increased number of connections? This will allow you identify which ports are in use and from which IPs, as well as the program using the socket.
Sorry for not getting back to you about this, were you able to collect the netstat? It would be interesting to see what state the connections are in.
If possible could you upload the netstat output that you collect after reproducing the issue?
Obviously feel free to sanitise any IPs if you think that they are identifying, although try to keep the same IP sanitised to the same name so it’s easy to match up connections (ideally you could mark the client IP).
Well it looks as though the connections displayed in the UI are legitimate!
It seems as though a large number of clients (from different IPs assuming that you didn’t randomise the last 3 bits of the foreign address) are connecting to the node, specifically via port 11210 which is the data access port.
The issue of the number of connections constantly increasing might not be a server-side issue (your original value of 70 could be normal considering all of the inter-node connections), but rather a client-side one.
What SDK do you use for your clients?
Would it be possible for you to share your connection + data access code for Couchbase, it may be that you are unexpectedly recreating connections.
Perhaps a little more info about your application setup would be useful here too!