Make shuffling of cluster urls work in tests

[kjk]

When setting SHUFFLE_CLUSTER_NODES to true, we get e.g. TestLoad to fail.

I start 3 server processes locally:

Started server 'C:\Users\kjk\src\ravendb-go-client\RavenDB\Server\Raven.Server.exe --ServerUrl=http://127.0.0.1:0 --ServerUrl.Tcp=tcp://127.0.0.1:38881 --RunInMemory=true --License.Eula.Accepted=true --Setup.Mode=None --Testing.ParentProcessId=21296' on port 'http://127.0.0.1:8048' pid: 19484
Started server 'C:\Users\kjk\src\ravendb-go-client\RavenDB\Server\Raven.Server.exe --ServerUrl=http://127.0.0.1:0 --ServerUrl.Tcp=tcp://127.0.0.1:38882 --RunInMemory=true --License.Eula.Accepted=true --Setup.Mode=None --Testing.ParentProcessId=21296' on port 'http://127.0.0.1:8051' pid: 19780
Started server 'C:\Users\kjk\src\ravendb-go-client\RavenDB\Server\Raven.Server.exe --ServerUrl=http://127.0.0.1:0 --ServerUrl.Tcp=tcp://127.0.0.1:38883 --RunInMemory=true --License.Eula.Accepted=true --Setup.Mode=None --Testing.ParentProcessId=21296' on port 'http://127.0.0.1:8053' pid: 20096

I create a cluster by adding 2nd and 3rd server via /admin/cluster/node as per https://ravendb.net/docs/article-page/4.1/csharp/server/clustering/cluster-api#bootstrap-cluster

PUT http://127.0.0.1:8048/admin/cluster/node?url=http%3A%2F%2F127.0.0.1%3A8051
Added node to cluster with 'http://127.0.0.1:8048/admin/cluster/node?url=http%3A%2F%2F127.0.0.1%3A8051', status code: 204
PUT http://127.0.0.1:8048/admin/cluster/node?url=http%3A%2F%2F127.0.0.1%3A8053
Added node to cluster with 'http://127.0.0.1:8048/admin/cluster/node?url=http%3A%2F%2F127.0.0.1%3A8053', status code: 204

Not sure if that's enough. What is /admin/cluster/bootstrap ? Do I need to wait for raft consensu to emerge before executing things against the cluster?

I then create a database, with replication factor 1 always on the first node, which should be the leader.

PUT http://127.0.0.1:8048/admin/databases?name=test_db_1&replicationFactor=1

Should I bump replication factor?

But when I shuffle urls for the new store / database, when I try to get topology for the database, it fails if executed against a server different than the first:

GET http://127.0.0.1:8053/topology?name=test_db_1
=========== 4:
GET /topology?name=test_db_1 HTTP/1.1
Host: 127.0.0.1:8053
Client-Configuration-Etag: "0"
Raven-Client-Version: 4.0.0
Topology-Etag: "0"
User-Agent: ravendb-go-client/4.0.0

--------
HTTP/1.1 503 Service Unavailable
Transfer-Encoding: chunked
Content-Type: application/json; charset=utf-8
Database-Missing: test_db_1
Date: Wed, 20 Feb 2019 23:25:49 GMT
Raven-Server-Version: 4.1.5.41
Server: Kestrel
Vary: Accept-Encoding

Response body 60 bytes:
{
  "Message": "Database test_db_1 wasn't found",
  "Type": "Error"
}

At which point I don't know:

• should this work and my cluster setup is incorrect? Do I need to wait somehow for "create database" command to finish replicating?
• is the problem that the code should then pick a different server and re-issue the command?