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
Provide an easy way to get all net.Server connections #48433
Currently, there seems no way to get all net.Server connections with the public API.
net.server.getConnections((err, count) => {...}) only returns the count of the connection, I can't get each connection within its callback. I tried for (let socket of server.connections) {...} inside the callback, and got the error TypeError: server.connections is not iterable.
I have to store each socket myself on Event: 'connection' callback, keep an array of those socket objects.
What is the feature you are proposing to solve the problem?
Inside net.server.getConnections((err, count) => {...}) callback, in addition to returning connection count, also return all the connections.
A similar feature was originally proposed in #42812, but I consider it an anti-pattern in Node.js (see #42812 (review)):
Constructing potentially large lists of resources generally feels like an anti-pattern in Node.js, and unless the application code is entirely synchronous, it is prone to race conditions.
Another reason not to do this is that it would be incongruent with server.getConnections()1. That method reports connections handled by worker processes but the feature being requested here can't do that because such connections are gone.
1 Which, for the record, I think is of questionable design too.
Honestly I can't imagine what current getConnections() could be useful for at all. Anyway everyone has to implement the list of connected sockets manually to be able to close the server fully. If an app serves millions of clients, the list would only occupy several MBs in memory which is nothing compared to all the allocated socket objects. Probably it's the lookups in containers with millions of items that are problem?
I can't imagine what current getConnections() could be useful for at all.
Especially @bnoordhuis said "such connections are gone", although I don't get why. If some connections are gone, what is the point of using count in its callback?
I assume getConnections() to count connections is mostly used as a metric in practice, e.g., with Prometheus or so.
And I believe what @bnoordhuis meant is that the primary process in a cluster does not have access to all connection objects because some exist in the address space / JS heap of other processes.
There has been no activity on this feature request for 5 months and it is unlikely to be implemented. It will be closed 6 months after the last non-automated comment.
There has been no activity on this feature request and it is being closed. If you feel closing this issue is not the right thing to do, please leave a comment.
Activity
tniessen commentedon Jun 13, 2023
A similar feature was originally proposed in #42812, but I consider it an anti-pattern in Node.js (see #42812 (review)):
bnoordhuis commentedon Jun 14, 2023
Another reason not to do this is that it would be incongruent with
server.getConnections()1. That method reports connections handled by worker processes but the feature being requested here can't do that because such connections are gone.1 Which, for the record, I think is of questionable design too.
Antonius-S commentedon Jun 15, 2023
Honestly I can't imagine what current getConnections() could be useful for at all. Anyway everyone has to implement the list of connected sockets manually to be able to close the server fully. If an app serves millions of clients, the list would only occupy several MBs in memory which is nothing compared to all the allocated socket objects. Probably it's the lookups in containers with millions of items that are problem?
qiulang commentedon Jun 17, 2023
@tniessen @bnoordhuis If it weren't for
net.getConnections()I won't ask this feature request.As @Antonius-S said
Especially @bnoordhuis said "such connections are gone", although I don't get why. If some connections are gone, what is the point of using count in its callback?
tniessen commentedon Jun 18, 2023
I assume
getConnections()to count connections is mostly used as a metric in practice, e.g., with Prometheus or so.And I believe what @bnoordhuis meant is that the primary process in a cluster does not have access to all connection objects because some exist in the address space / JS heap of other processes.
github-actions commentedon Dec 16, 2023
There has been no activity on this feature request for 5 months and it is unlikely to be implemented. It will be closed 6 months after the last non-automated comment.
For more information on how the project manages feature requests, please consult the feature request management document.
github-actions commentedon Jan 15, 2024
There has been no activity on this feature request and it is being closed. If you feel closing this issue is not the right thing to do, please leave a comment.
For more information on how the project manages feature requests, please consult the feature request management document.