You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
How does the client record the task IDs that have been sent #909
There is a scenario where the client initiates event subscription to the server and maintains a long connection through SSE. When the SSE is disconnected due to network fluctuations, the client needs to re-subscribe.
But at this point, how can the client know the task ID of the subscription?
Can it only be achieved by persisting the context ID or the task ID?
Hi, yes you would need to do something in the client application to track the task IDs, or perhaps you could list the tasks. But I think listing would mean you need to know the ones you want anyway, which again implies some tracking would be needed
Hi, yes you would need to do something in the client application to track the task IDs, or perhaps you could list the tasks. But I think listing would mean you need to know the ones you want anyway, which again implies some tracking would be needed
It is uncertain whether information other than contextId should be introduced in future protocols to identify clients.
However, there is another scenario where two pods automatically initiate subscription tasks to the server upon startup.
In this case, the issue of multiple instances arises, making it difficult to assign a client ID. If each client were to perform leader election, it would be a cumbersome process.
Activity
kabir commentedon Jun 3, 2026
Hi, yes you would need to do something in the client application to track the task IDs, or perhaps you could list the tasks. But I think listing would mean you need to know the ones you want anyway, which again implies some tracking would be needed
Wesley273 commentedon Jun 4, 2026
It is uncertain whether information other than contextId should be introduced in future protocols to identify clients.
However, there is another scenario where two pods automatically initiate subscription tasks to the server upon startup.
In this case, the issue of multiple instances arises, making it difficult to assign a client ID. If each client were to perform leader election, it would be a cumbersome process.