Re: connect(2) for UDP
Peter Deacon <[email protected]>
| Newsgroups | gmane.comp.db.tds.freetds |
|---|---|
| Message-ID | <alpine.WNT.2.00.0812172209440.1988@SMURF> |
On Wed, 17 Dec 2008, James K. Lowden wrote: >> Good point -- we had a server with three IPs on one network, so it could >> have 3 web server instances each listening on port 80. The source >> address of the SNMP UDP packets would end up being round-robined. > Is this a theory, or observed behavior? > If I send a message to a port at an address, I more or less expect an > answer back from the same place. Even if each machine in the cluster > listens to 1434, the internal cluster logic has to be simpler if each > machine answers its own queries.... UDP has no concept of connection. Each message stands alone. If the server binds to all interfaces and there are multiple interfaces capable of reaching the client the source address of messages from the server should be concidered random/determined by the routing system. Servers that need to make sure the IPs match up in both directions for security checking..etc will typically separatly bind to each local interface individually to ensure the source address from the client perspective are deterministic. A quick netstat should show how MSSQL is binding and what behavior to expect. regards, Peter