/Distributed-Auction-System

Primary LanguageGo

Distributed Auction System

BSDISYS1KU-20222, mandatory handin 5

How to run

  1. Run server.go & client.go in separate terminals - the order is very important, the servers must be running first.

In the source code for server.go - you will find a const BASEPORT, this is the port that the servers will incrementally use. You do not need to supply a port through commandline, if the baseport is not available - a server will increment and try again.

In the source code for client.go - you will find const BASEPORT also (this needs to match server.go), and const REPLICAS. While it is not necessary to set const REPLICAS to the same amount that of server instances you've started - it does make sense to do, since it prevents having to wait for timeouts to finish.

```console
$ go run .\server\server.go
$ go run .\server\server.go
$ go run .\server\server.go
$ go run .\server\server.go
$ go run .\client\client.go *
```

The client needs a parameter of uint32 - this is the ID of the client when bidding. There is nothing that checks for whether or not an ID is in use, so just make sure you do not use duplicates

  1. If you want to stress the system, but you are not able to manually input at the speed you want - you can configure AUTOCLIENT, MIN_DELAY, and MAX_DELAY in client.go.

These will do the equivalent of fuzzing the replicas - just spamming whatever they generate.

Stuff that might go wrong

We doubt that you will encounter any of this, since we are using localhost & our PC's are not good (to put it nicely) - however, now you know what to try if you have any of the issues :)

  1. If you are encountering different results than we've shown in our logs & report - your net might be less stable than ours. In that case, find the function DelayedUnlock() (line 230) in server.go and up the sleep untill you attain stable results.

  2. Clients do not find the servers - timing out on initial dial. You can try to up the value on line 67 in client.go

  3. Replica failing crashes client - this should only happen if you are only running one replica, and you crash it. We did not do any kind of error handling for if there are no valid replicas - because the assignment did not have this as part of the scope :)