1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
|
// -*- Mode: Go; indent-tabs-mode: t -*-
/*
* Copyright (C) 2025 Canonical Ltd
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 3 as
* published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
*/
package assemblestate
import (
"errors"
"sort"
"time"
"github.com/snapcore/snapd/cluster/assemblestate/bimap"
"github.com/snapcore/snapd/cluster/assemblestate/intset"
"github.com/snapcore/snapd/randutil"
)
// RouteSelector keeps track of which routes we've seen and helps pick which
// peers to publish routes to. While transparent in the interface, a
// RouteSelector is provided a source to query which peers are identified.
type RouteSelector interface {
// AddAuthoritativeRoute records a route from this local node to the given
// [DeviceToken]. This route will be published to the given peer, regardless
// of our knowledge of that peer's identity.
//
// We treat these routes slightly differently because of how the protocol
// defines what peers should do upon receipt of routes. If we send a route
// to our peer that includes devices that they do not yet know, it is
// expected that they can request identifying information for those devices
// from us. Since a peer always knows about itself, it is safe for us to
// publish an authoritative route to the peer at the route's destination.
AddAuthoritativeRoute(to DeviceToken, via string)
// RecordRoutes records a set of routes from the given [RDT]. If a route's
// origin and destination both are reported as known by the selector's
// knowledge of device identity, then those routes can be considered for
// publication.
RecordRoutes(from DeviceToken, r Routes) (int, int, error)
// VerifyRoutes re-calculates which routes are available for publication.
// For all routes that are already known, they will be marked as available
// for publication if the selector knows the identity of the route's origin
// and destination devices.
//
// This method should be called whenever the RouteSelector's source of
// knowledge of device identities changes.
VerifyRoutes()
// Select selects a subset of routes that the specified peer needs to receive.
//
// Returns the routes to send, an acknowledgment function that should be
// called after successful transmission, and whether routes were selected.
// The ack function must be called once the selected routes are published
// so that they will not be selected for publication again.
Select(to DeviceToken, count int) (routes Routes, ack func(), ok bool)
// Routes returns all routes that are currently valid for publication.
Routes() Routes
}
// PrioritySelector implements [RouteSelector].
//
// This implementation randomly selects one peer for route publication.
// Additionally, we prioritize routes that this local node has witnessed and
// published the smallest number of times.
type PrioritySelector struct {
// self is this node's RDT.
self DeviceToken
// rng is the random number generator that is used for peer selection.
rng *randutil.Rand
// identified is a function that determines whether a device is
// known/identified. This is provided by the caller, and the underlying data
// isn't managed by us.
identified func(DeviceToken) bool
// rdts keeps track of all RDTs that we've seen and maps them to a [peerID].
rdts *bimap.Bimap[DeviceToken, peerID]
// edges keeps track of all edges we've seen and maps them to an [edgeID].
edges *bimap.Bimap[edge, edgeID]
// addresses keeps track of all addresses we've seen and maps them to an
// [addrID].
addresses *bimap.Bimap[string, addrID]
// knownByPeers keeps track which routes each peer knows about. A route is
// considered known by a peer if either they have sent it to us, or we've
// sent it to them.
knownByPeers map[peerID]*intset.IntSet[edgeID]
// edgeSources keeps track of how many unique peers we've seen an edge from
// and sent an edge to. This helps us prioritize which routes to send to our
// peers.
edgeSources map[edgeID]int
// verifiedEdges keeps track of which edges are safe to publish. These routes
// only include RDTs for devices that are reported as identified by our
// caller.
verifiedEdges *intset.IntSet[edgeID]
// authoritative keeps track of the set of edges from this local node to
// each other peer. Each of these can be safely sent to the destination node
// in the edge.
authoritative map[DeviceToken]edgeID
}
type Identifier = func(DeviceToken) bool
func NewPrioritySelector(
self DeviceToken,
source randutil.Source,
identified func(DeviceToken) bool,
) *PrioritySelector {
if source == nil {
source = randutil.NewSource(time.Now().UnixNano())
}
return &PrioritySelector{
self: self,
rng: randutil.New(source),
identified: identified,
rdts: bimap.New[DeviceToken, peerID](),
edges: bimap.New[edge, edgeID](),
addresses: bimap.New[string, addrID](),
verifiedEdges: &intset.IntSet[edgeID]{},
knownByPeers: make(map[peerID]*intset.IntSet[edgeID]),
edgeSources: make(map[edgeID]int),
authoritative: make(map[DeviceToken]edgeID),
}
}
type (
// peerID is an opaque identifier that represents a peer. This is used to
// intern our strings to limit memory usage.
peerID int
// edgeID is an opaque identifier that represents an edge. This is used to
// intern our strings to limit memory usage.
edgeID int
// addrID is an opaque identifier that represents an address. This is used
// to intern our strings to limit memory usage.
addrID int
)
type edge struct {
from, to peerID
via addrID
}
func (p *PrioritySelector) peerID(rdt DeviceToken) peerID {
if pid, ok := p.rdts.IndexOf(rdt); ok {
return pid
}
pid := p.rdts.Add(rdt)
p.knownByPeers[pid] = &intset.IntSet[edgeID]{}
return pid
}
func (p *PrioritySelector) addrID(a string) addrID {
return p.addresses.Add(a)
}
func (p *PrioritySelector) edgeID(e edge) (id edgeID, existed bool) {
if eid, ok := p.edges.IndexOf(e); ok {
return eid, true
}
return p.edges.Add(e), false
}
// AddAuthoritativeRoute informs the selector of an authoritative route from
// this local node to the given peer. This route can safely be published to the
// given peer, regardless of our knowledge of that peer's identity.
func (p *PrioritySelector) AddAuthoritativeRoute(to DeviceToken, via string) {
eid, _ := p.edgeID(edge{
from: p.peerID(p.self),
to: p.peerID(to),
via: p.addrID(via),
})
p.authoritative[to] = eid
}
// RecordRoutes records all give routes and marks them as known to the given
// [DeviceToken]. The selector's identified function is used to verify routes
// and mark them as safe to publish if all we know all devices involved in a
// route.
func (p *PrioritySelector) RecordRoutes(source DeviceToken, r Routes) (added int, total int, err error) {
pid := p.peerID(source)
if len(r.Routes)%3 != 0 {
return 0, 0, errors.New("length of routes list must be a multiple of three")
}
// r.Routes is a slice triplets, where each triplet represents a route
// between two devices in our cluster. the values in triplet are indexes
// into the other slices in the [Routes] message.
// - r.Routes[n] is an index into r.Devices, representing the origin of the
// route
// - r.Routes[n+1] is an index into r.Devices, representing the
// destination of the route
// - r.Routes[n+1] is an index into r.Addresses, representing the address
// that r.Routes[n] used to reach r.Routes[n+1]
for i := 0; i+2 < len(r.Routes); i += 3 {
if r.Routes[i] < 0 || r.Routes[i+1] < 0 || r.Routes[i+2] < 0 {
return 0, 0, errors.New("route contains negative index")
}
if r.Routes[i] >= len(r.Devices) || r.Routes[i+1] >= len(r.Devices) || r.Routes[i+2] >= len(r.Addresses) {
return 0, 0, errors.New("route index exceeds available devices or addresses")
}
fromRDT := r.Devices[r.Routes[i]]
toRDT := r.Devices[r.Routes[i+1]]
fromID := p.peerID(fromRDT)
toID := p.peerID(toRDT)
viaID := p.addrID(r.Addresses[r.Routes[i+2]])
eid, existed := p.edgeID(edge{
from: fromID,
to: toID,
via: viaID,
})
if !existed {
added++
}
// if we aren't aware that this peer knows about this route already,
// increment our counter of sources for that edge
if !p.knownByPeers[pid].Contains(eid) {
p.edgeSources[eid]++
}
// record that the peer who sent this Routes message knows about this
// edge
p.knownByPeers[pid].Add(eid)
// if we have the identities of both the from and to devices, then we
// know can verify this route. verified routes can published to our
// peers.
if p.identified(fromRDT) && p.identified(toRDT) {
p.verifiedEdges.Add(eid)
}
}
return added, len(p.edges.Values()), nil
}
// VerifyRoutes uses the selector's identified function to mark any routes that
// involve devices that we know as safe to publish.
func (p *PrioritySelector) VerifyRoutes() {
for eid, edge := range p.edges.Values() {
fromRDT := p.rdts.Value(edge.from)
toRDT := p.rdts.Value(edge.to)
if p.identified(fromRDT) && p.identified(toRDT) {
p.verifiedEdges.Add(edgeID(eid))
}
}
}
// Select selects n routes that should be published to the specified peer.
//
// We prioritize routes that the local node has witnessed and published fewer
// times.
func (p *PrioritySelector) Select(to DeviceToken, n int) (routes Routes, ack func(), ok bool) {
selected, exists := p.rdts.IndexOf(to)
if !exists {
return Routes{}, nil, false
}
if to == p.self {
return Routes{}, nil, false
}
peerKnown := p.knownByPeers[selected]
unknown := p.verifiedEdges.Diff(peerKnown)
// max possible needed size is all unknown routes + the route from this
// local node to the destination peer
edgesToSend := make([]edgeID, 0, unknown.Count()+1)
// only consider routes that we don't think that this peer knows about
unknown.Range(func(eid edgeID) bool {
edgesToSend = append(edgesToSend, eid)
return true
})
// prioritize routes that we think fewer peers know about. this takes into
// account copies of routes we've seen from our peers and copies that we've
// sent to our peers
sort.Slice(edgesToSend, func(i, j int) bool { return p.edgeSources[edgesToSend[i]] < p.edgeSources[edgesToSend[j]] })
// discard any edges causing us to exceed the given threshold. thus, we pick
// the n least frequently seen routes.
edgesToSend = edgesToSend[:min(len(edgesToSend), n)]
// if we have an authoritative route for this peer, make sure to include it
if eid, ok := p.authoritative[to]; ok && !peerKnown.Contains(eid) {
edgesToSend = append(edgesToSend, eid)
}
if len(edgesToSend) == 0 {
return Routes{}, nil, false
}
routes = p.edgesToRoutes(edgesToSend)
// create ack function that updates source counts when called
ack = func() {
for _, eid := range edgesToSend {
// the peer might know about the route already since selection time.
// if that has happened, then we don't want to double count that
// peer as a source.
if !peerKnown.Contains(eid) {
peerKnown.Add(eid)
p.edgeSources[eid]++
}
}
}
return routes, ack, true
}
// Routes returns routes that we've seen for which both peers in the route
// identified.
func (p *PrioritySelector) Routes() Routes {
eids := p.verifiedEdges.All()
devs := make([]string, 0, len(p.rdts.Values()))
devIndexes := make(map[string]int, cap(devs))
addrs := make([]string, 0, len(p.addresses.Values()))
addrIndexes := make(map[string]int, cap(addrs))
// here we build a slice of devices and address that are used in the routes.
// we want the slices to be sorted for output consistency, so we also build
// a mapping of values to indexes in these slices. this first pass just
// fills the maps with placeholders (-1) to track seen values, later we
// actually assign indexes to the values after sorting.
for _, eid := range eids {
edge := p.edges.Value(eid)
to := string(p.rdts.Value(edge.to))
if _, ok := devIndexes[to]; !ok {
devs = append(devs, to)
devIndexes[to] = -1
}
from := string(p.rdts.Value(edge.from))
if _, ok := devIndexes[from]; !ok {
devs = append(devs, from)
devIndexes[from] = -1
}
via := p.addresses.Value(edge.via)
if _, ok := addrIndexes[via]; !ok {
addrs = append(addrs, via)
addrIndexes[via] = -1
}
}
// these checks should be impossible
if len(devs) != len(devIndexes) {
panic("internal error: invalid device count when exporting routes")
}
if len(addrs) != len(addrIndexes) {
panic("internal error: invalid address count when exporting routes")
}
sort.Strings(devs)
for i, d := range devs {
devIndexes[d] = i
}
sort.Strings(addrs)
for i, a := range addrs {
addrIndexes[a] = i
}
sort.Slice(eids, func(i, j int) bool {
a, b := p.edges.Value(eids[i]), p.edges.Value(eids[j])
if p.rdts.Value(a.from) != p.rdts.Value(b.from) {
return p.rdts.Value(a.from) < p.rdts.Value(b.from)
}
if p.rdts.Value(a.to) != p.rdts.Value(b.to) {
return p.rdts.Value(a.to) < p.rdts.Value(b.to)
}
return p.addresses.Value(a.via) < p.addresses.Value(b.via)
})
routes := make([]int, 0, len(eids)*3)
for _, eid := range eids {
edge := p.edges.Value(eid)
from := devIndexes[string(p.rdts.Value(edge.from))]
to := devIndexes[string(p.rdts.Value(edge.to))]
via := addrIndexes[string(p.addresses.Value(edge.via))]
routes = append(routes,
from,
to,
via,
)
}
converted := make([]DeviceToken, 0, len(devs))
for _, d := range devs {
converted = append(converted, DeviceToken(d))
}
return Routes{
Devices: converted,
Addresses: addrs,
Routes: routes,
}
}
func (p *PrioritySelector) edgesToRoutes(edges []edgeID) Routes {
rdts := bimap.New[DeviceToken, int]()
addrs := bimap.New[string, int]()
routes := make([]int, 0, len(edges)*3)
for _, eid := range edges {
edge := p.edges.Value(eid)
from := p.rdts.Value(edge.from)
to := p.rdts.Value(edge.to)
address := p.addresses.Value(edge.via)
routes = append(routes,
rdts.Add(from),
rdts.Add(to),
addrs.Add(address),
)
}
return Routes{
Devices: rdts.Values(),
Addresses: addrs.Values(),
Routes: routes,
}
}
// min returns the smaller of the two given values.
//
// TODO: remove once we are on go>=1.21
func min(x int, y int) int {
if x < y {
return x
}
return y
}
|