Brad Fitzpatrick
f87a1b1a82
Every packet the server relayed allocated a fresh []byte for its payload in recvPacket or recvForwardPacket and dropped it once the destination's sendLoop had written it. On one busy server, this was observed allocating about 160 MB/sec of short-lived garbage, and GC plus malloc were about 5% of the process CPU profile. Instead, take payload buffers from a size-classed sync.Pool on the Server, with power-of-two classes from 1 KiB up to derp.MaxPacketSize, and return them once the packet has been written, forwarded, or dropped. sync.Pool holds nothing per connection and is trimmed by the GC, so idle clients pin no memory; only packets actually in flight hold a buffer. A compile-time assertion ties the largest size class to derp.MaxPacketSize, and the get and put helpers panic on sizes outside the pool's classes rather than indexing past it. Because the memory is now reused, PacketForwarder implementations must not retain the payload after ForwardPacket returns. Make that explicit in the signature: the payload is passed as a new derp.LoanedBytes value, which exposes only Len, WriteTo, and Clone, so an implementation has to copy to keep it. derp.Client and derphttp.Client, the real implementations, already wrote it out synchronously; the test-only channelFwd now clones. BenchmarkSendRecv shows one fewer allocation per relayed packet and, for 1000-byte packets, B/op down from 1278 to 263. ns/op on the loopback benchmarks is dominated by syscalls and is unchanged within noise. Updates #21064 Change-Id: Ie40c82388ddb5d22f75fa828749b53fcaba9adde Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
DERP
This directory (and subdirectories) contain the DERP code. The server itself is
in ../cmd/derper.
DERP is a packet relay system (client and servers) where peers are addressed using WireGuard public keys instead of IP addresses.
It relays two types of packets:
-
"Disco" discovery messages (see
../disco) as the a side channel during NAT traversal. -
Encrypted WireGuard packets as the fallback of last resort when UDP is blocked or NAT traversal fails.
DERP Map
Each client receives a "DERP Map" from the coordination server describing the DERP servers the client should try to use.
The client picks its home "DERP home" based on latency. This is done to keep costs low by avoid using cloud load balancers (pricey) or anycast, which would necessarily require server-side routing between DERP regions.
Clients pick their DERP home and report it to the coordination server which shares it to all the peers in the tailnet. When a peer wants to send a packet and it doesn't already have a WireGuard session open, it sends disco messages (some direct, and some over DERP), trying to do the NAT traversal. The client will make connections to multiple DERP regions as needed. Only the DERP home region connection needs to be alive forever.
DERP Regions
Tailscale runs 1 or more DERP nodes (instances of cmd/derper) in various
geographic regions to make sure users have low latency to their DERP home.
Regions generally have multiple nodes per region "meshed" (routing to each
other) together for redundancy: it allows for cloud failures or upgrades without
kicking users out to a higher latency region. Instead, clients will reconnect to
the next node in the region. Each node in the region is required to be meshed
with every other node in the region and forward packets to the other nodes in
the region. Packets are forwarded only one hop within the region. There is no
routing between regions. The assumption is that the mesh TCP connections are
over a VPC that's very fast, low latency, and not charged per byte. The
coordination server assigns the list of nodes in a region as a function of the
tailnet, so all nodes within a tailnet should generally be on the same node and
not require forwarding. Only after a failure do clients of a particular tailnet
get split between nodes in a region and require inter-node forwarding. But over
time it balances back out. There's also an admin-only DERP frame type to force
close the TCP connection of a particular client to force them to reconnect to
their primary if the operator wants to force things to balance out sooner.
(Using the (*derphttp.Client).ClosePeer method, as used by Tailscale's
internal rarely-used cmd/derpprune maintenance tool)
We generally run a minimum of three nodes in a region not for quorum reasons (there's no voting) but just because two is too uncomfortably few for cascading failure reasons: if you're running two nodes at 51% load (CPU, memory, etc) and then one fails, that makes the second one fail. With three or more nodes, you can run each node a bit hotter.