If i wanted to start an agent on k8 that would launch no new flows, but continue to monitor any existing flows currently running would either of these solutions work? • Start the agent with

--limit=0

• Start the agent with no

--work-queue

parameter My use case is doing a live failover between two k8 clusters (A and B), where I’d want all new workloads being sent to cluster B, while still allowing active workloads on A to tapper off.

Hi @Josh Paulin, currently we do not offer a way to spin up an agent without including a work queue, but to achieve your pattern of monitoring currently running flows, you can programmatically create additional agents that are spun up whenever flows are in the running state. Additionally, you can outsource these logs to a file as well to be tracked in a similar fashion via rest api calls.

Not sure this is really answering my question, but maybe my use case isn’t clear. Assume I have a k8 cluster

A

that is running an agent and running flows. I need to do a failover to a second cluster

B

, such that all new flow runs get run on

B

, while flows runs that are running on

A

are allowed to gracefully complete. If #2 wouldn’t work (tbh I wasn’t expecting it to), and I able to restart the agent on

A

with a

--limit=0

so that it will gracefully draw down any active flows on the cluster