satellite/orders: send IPs for graceful exit
Storage nodes undergoing Graceful Exit have up to now been receiving hostnames for all other storage nodes they need to contact when transferring pieces. This adds up to a lot of DNS lookups, which apparently overwhelm some home routers. There does not seem to be any need for us to send hostnames for graceful exit as opposed to IP addresses; we already use IP addresses (as given by the last_ip_port column in the nodes table) for all the GET and PUT orders we send out. This change causes IP addresses to be used instead. I started trying to construct a test to ensure that the behavior changed, but it was rabbit-holing, so I've begun to feel that maybe this change doesn't require one; it is a very simple change, and very much of the same nature as what we already do for IPs in CreateGetOrderLimits and CreatePutOrderLimits (and others). Change-Id: Ib2b5ffe7a9310e9cdbe7464450cc7c934fa229a1
This commit is contained in:
parent
c55c23f81f
commit
8616fc146d
@ -566,7 +566,11 @@ func (service *Service) CreateGracefulExitPutOrderLimit(ctx context.Context, buc
|
||||
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
|
||||
}
|
||||
|
||||
nodeURL := storj.NodeURL{ID: nodeID, Address: node.Address.Address}
|
||||
address := node.Address.Address
|
||||
if node.LastIPPort != "" {
|
||||
address = node.LastIPPort
|
||||
}
|
||||
nodeURL := storj.NodeURL{ID: nodeID, Address: address}
|
||||
limit, err = signer.Sign(ctx, nodeURL, pieceNum)
|
||||
if err != nil {
|
||||
return nil, storj.PiecePrivateKey{}, Error.Wrap(err)
|
||||
|
Loading…
Reference in New Issue
Block a user