CDN Observability
Full Recipe¶
Shared by: Allan Konar
Real-time computation of CDN cache node efficiency from pseudonymized Fastly CDN logs, with graph association of each log entry to serving PoP, cache server, client, client ASN, asset and origin to identify potential root cause of issues.
Full CDN Observability Recipe
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 |
|
Scenario¶
Pseudonymized CDN log data is imported from a JSON file (cdn_data_50k.json
) via a file ingest, and nodes are manifested for the elements associated with each event (e.g., client, server, pop, etc.).
Each of the manifested nodes increment counters to track the number of cache hits and misses at each level (e.g., source ASN, server, pop, etc.). Selecting any node allows you to query at each level to identify potential root cause of poor performance.
A standing query is defined to match consecutive cache misses within a configurable fixed period of time for the purpose of alerting.
Sample Data¶
Download the sample data to the same directory where Quine will be run.
cdn_data_50k.json
- https://that.re/cdn_data_50k
How it Works¶
The recipe reads observations from the sample data file using ingest streams to manifest a graph in Quine. An ingest stream is configured to process the data file, containing Cypher that parses the log entries, manifests nodes, and relates them to each other in the graph.
The log entries take the form of:
{
"backend_ip": "157.52.79.52",
"backend_ttlb": 73.206,
"business_unit": "68ae725c3fd8d6831735753269a727c9ce05baae6715bb0191ddb7f6d67842bd",
"bytes_in": 377,
"bytes_out": 902,
"cached": false,
"cache_shield": "false",
"cache_status": "MISS-CLUSTER",
"client_asn": 7922,
"client_geo_country": "US",
"client_ip": "1localhost",
"client_ttfb": 73.198,
"environment": "prod",
"failover_status": "",
"forward_for": "",
"host": "682d313399617e3d194679e8422a6a5f2666b60c54ab6eead5908040443c793f",
"if_modified_since": "",
"if_none_match": "",
"if_unmodified_since": "",
"method": "GET",
"path": "/flavi7d79/master/flavi7d79_6.m3u8",
"pop": "SJC",
"range_request": "(null)",
"range_response": "",
"request_id": "cache-sjc10025-SJC-2614091026",
"restarts": 0,
"retrans": 0,
"role": "edge",
"rtt_msecs": 28,
"server_id": "cache-sjc10025-SJC",
"server_ip": "2a04:4e42:a::645",
"server_ttlb": 73.261,
"shield_failover": "",
"stream": 0,
"status_code": 200,
"timestamp": "2020-07-14 22:55:31.729734",
"user_agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/83.0.4103.116 Safari/537.36",
"workflow": "f2757f5a18302320de05caa509ac98370b6a9cec",
"origin_request_id": "",
"query": ""
}
INGEST-1 processes the cdn_data_50k.json
file:
- type: FileIngest
path: cdn_data_50k.json
format:
type: CypherJson
query: |-
MATCH (event), (client), (asset), (asn), (server), (pop), (origin), (clientGeo)
WHERE $that.cache_status IS NOT NULL
AND id(event) = idFrom('event', $that.timestamp, $that.request_id)
AND id(client) = idFrom('client', $that.client_ip, $that.business_unit)
AND id(asset) = idFrom('asset', $that.path)
AND id(asn) = idFrom('asn', toString($that.client_asn))
AND id(server) = idFrom('server', $that.pop, $that.server_id)
AND id(pop) = idFrom('pop', $that.pop)
AND id(origin) = idFrom('origin', $that.backend_ip)
AND id(clientGeo) = idFrom('clientGeo', $that.client_geo_country)
////////////////////////////////////////
//Bucketing for HITs and MISSes counters
////////////////////////////////////////
// RegEx deets here: https://regex101.com/r/uP0KMm/1
WITH *, text.regexFirstMatch($that.cache_status, '(HIT|MISS(?!.*HIT)).*') AS hmp WHERE hmp[1] IS NOT NULL
////////////////////////////////////////
// Bucketing for node type counters
////////////////////////////////////////
CALL incrementCounter(client, "count",1) YIELD count AS clientCount
CALL incrementCounter(client, toLower(hmp[1]),1) YIELD count AS clientHitMissCount
CALL incrementCounter(asset, "count",1) YIELD count AS assetCount
CALL incrementCounter(asset, toLower(hmp[1]),1) YIELD count AS assetHitMissCount
CALL incrementCounter(asn, "count",1) YIELD count AS asnCount
CALL incrementCounter(asn, toLower(hmp[1]),1) YIELD count AS asnHitMissCount
CALL incrementCounter(server, "count",1) YIELD count AS serverCount
CALL incrementCounter(server, toLower(hmp[1]),1) YIELD count AS serverHitMissCount
CALL incrementCounter(pop, "count",1) YIELD count AS popCount
CALL incrementCounter(pop, toLower(hmp[1]),1) YIELD count AS popHitMissCount
CALL incrementCounter(clientGeo, "count",1) YIELD count AS clientGeoCount
CALL incrementCounter(clientGeo, toLower(hmp[1]),1) YIELD count AS clientGeoHitMissCount
CALL incrementCounter(origin, "count",1) YIELD count AS originGeoCount
CALL incrementCounter(origin, toLower(hmp[1]),1) YIELD count AS originGeoHitMissCount
////////////////////////////////////////////////////////
// Event
////////////////////////////////////////////////////////
SET event = $that,
event.cache_class = hmp[1],
event: event
////////////////////////////////////////////////////////
// Origin
////////////////////////////////////////////////////////
SET origin.backend_ip = $that.backend_ip,
origin: origin
////////////////////////////////////////////////////////
// Client
////////////////////////////////////////////////////////
SET client.client_geo_country = $that.client_geo_country,
client.client_ip = $that.client_ip,
client.user_agent = $that.user_agent,
client: client
// Extract Browser and Version
// RegEx here: https://regex101.com/r/T0MThZ/2
WITH *, text.regexFirstMatch($that.user_agent, '\\((.*?)\\)(\\s|$)|(.*?)\\/(.*?)(\\s|$)') AS cb
SET client.browser = cb[3],
client.browserVer = cb[4],
client.first_seen = coll.min([$that.timestamp, coalesce(client.first_seen, $that.timestamp)]),
client.last_seen = coll.max([$that.timestamp, coalesce(client.last_seen, $that.timestamp)])
////////////////////////////////////////////////////////
// Client Geo
////////////////////////////////////////////////////////
SET clientGeo.client_geo_country = $that.client_geo_country,
clientGeo: clientGeo
////////////////////////////////////////////////////////
// Asset
////////////////////////////////////////////////////////
// RegEx here: https://regex101.com/r/tB8cd4/1
WITH *, text.regexFirstMatch($that.path, '^(.+\\/)([^\\/]+)$') AS ap
SET asset.path = ap[1],
asset.name = ap[2],
asset.full_path = $that.path,
asset.if_modified_since = coll.max([$that.timestamp, coalesce(asset.if_modified_since, $that.timestamp)]),
asset: asset
////////////////////////////////////////////////////////
// ASN
////////////////////////////////////////////////////////
SET asn.asn_id = toString($that.client_asn),
asn: asn
////////////////////////////////////////////////////////
// Server
////////////////////////////////////////////////////////
SET server.server_id = $that.server_id,
server.server_ip = $that.server_ip,
server.cache_shield = $that.cache_shield,
server.environment = $that.environment,
server.host = $that.host,
server.role = $that.role,
server.pop = $that.pop,
server: server
////////////////////////////////////////////////////////
// PoP
////////////////////////////////////////////////////////
SET pop.source = $that.pop,
pop.environment = $that.environment,
pop: pop
////////////////////////////////////////////////////////
// Create relationship between nodes
////////////////////////////////////////////////////////
CREATE (asset)<-[:REQUESTED]-(event)-[:REQUESTED_OVER]->(asn)-[:IN_CLIENT_GEO]->(clientGeo),
(origin)<-[:FROM]-(pop)<-[:WITHIN]-(server)<-[:TARGETED]-(event)<-[:ORIGINATED]-(client)
{
"type": "FileIngest",
"path": "cdn_data_50k.json",
"format": {
"type": "CypherJson",
"query": "MATCH (event), (client), (asset), (asn), (server), (pop), (origin), (clientGeo) WHERE $that.cache_status IS NOT NULL AND id(event) = idFrom('event', $that.timestamp, $that.request_id) AND id(client) = idFrom('client', $that.client_ip, $that.business_unit) AND id(asset) = idFrom('asset', $that.path) AND id(asn) = idFrom('asn', toString($that.client_asn)) AND id(server) = idFrom('server', $that.pop, $that.server_id) AND id(pop) = idFrom('pop', $that.pop) AND id(origin) = idFrom('origin', $that.backend_ip) AND id(clientGeo) = idFrom('clientGeo', $that.client_geo_country) WITH *, text.regexFirstMatch($that.cache_status, '(HIT|MISS(?!.*HIT)).*') AS hmp WHERE hmp[1] IS NOT NULL CALL incrementCounter(client, "count",1) YIELD count AS clientCount CALL incrementCounter(client, toLower(hmp[1]),1) YIELD count AS clientHitMissCount CALL incrementCounter(asset, "count",1) YIELD count AS assetCount CALL incrementCounter(asset, toLower(hmp[1]),1) YIELD count AS assetHitMissCount CALL incrementCounter(asn, "count",1) YIELD count AS asnCount CALL incrementCounter(asn, toLower(hmp[1]),1) YIELD count AS asnHitMissCount CALL incrementCounter(server, "count",1) YIELD count AS serverCount CALL incrementCounter(server, toLower(hmp[1]),1) YIELD count AS serverHitMissCount CALL incrementCounter(pop, "count",1) YIELD count AS popCount CALL incrementCounter(pop, toLower(hmp[1]),1) YIELD count AS popHitMissCount CALL incrementCounter(clientGeo, "count",1) YIELD count AS clientGeoCount CALL incrementCounter(clientGeo, toLower(hmp[1]),1) YIELD count AS clientGeoHitMissCount CALL incrementCounter(origin, "count",1) YIELD count AS originGeoCount CALL incrementCounter(origin, toLower(hmp[1]),1) YIELD count AS originGeoHitMissCount SET event = $that, event.cache_class = hmp[1], event: event SET origin.backend_ip = $that.backend_ip, origin: origin SET client.client_geo_country = $that.client_geo_country, client.client_ip = $that.client_ip, client.user_agent = $that.user_agent, client: client WITH *, text.regexFirstMatch($that.user_agent, '\\((.*?)\\)(\\s|$)|(.*?)\\/(.*?)(\\s|$)') AS cb SET client.browser = cb[3], client.browserVer = cb[4], client.first_seen = coll.min([$that.timestamp, coalesce(client.first_seen, $that.timestamp)]), client.last_seen = coll.max([$that.timestamp, coalesce(client.last_seen, $that.timestamp)]) SET clientGeo.client_geo_country = $that.client_geo_country, clientGeo: clientGeo WITH *, text.regexFirstMatch($that.path, '^(.+\\/)([^\\/]+)$') AS ap SET asset.path = ap[1], asset.name = ap[2], asset.full_path = $that.path, asset.if_modified_since = coll.max([$that.timestamp, coalesce(asset.if_modified_since, $that.timestamp)]), asset: asset SET asn.asn_id = toString($that.client_asn), asn: asn SET server.server_id = $that.server_id, server.server_ip = $that.server_ip, server.cache_shield = $that.cache_shield, server.environment = $that.environment, server.host = $that.host, server.role = $that.role, server.pop = $that.pop, server: server SET pop.source = $that.pop, pop.environment = $that.environment, pop: pop CREATE (asset)<-[:REQUESTED]-(event)-[:REQUESTED_OVER]->(asn)-[:IN_CLIENT_GEO]->(clientGeo),(origin)<-[:FROM]-(pop)<-[:WITHIN]-(server)<-[:TARGETED]-(event)<-[:ORIGINATED]-(client)
}
}
A standing query is configured to look for 10 consecutive cache MISS events involving the same server and asset pair within a defined duration.
- pattern:
type: Cypher
query: |-
MATCH (server1:server)<-[:TARGETED]-(event1 {cache_class:"MISS"})-[:REQUESTED]->(asset)<-[:REQUESTED]-(event2 {cache_class:"MISS"})-[:TARGETED]->(server2:server)
RETURN DISTINCT id(event1) AS event1
[
{
"pattern": {
"type": "Cypher",
"query": "MATCH (server1:server)<-[:TARGETED]-(event1 {cache_class:"MISS"})-[:REQUESTED]->(asset)<-[:REQUESTED]-(event2 {cache_class:"MISS"})-[:TARGETED]->(server2:server) RETURN DISTINCT id(event1) AS event1"
},
"outputs": {
"cacheMissAlert": {
"type": "CypherQuery",
"query": "MATCH (server1:server)<-[:TARGETED]-(event1 {cache_class:"MISS"})-[:REQUESTED]->(asset)<-[:REQUESTED]-(event2 {cache_class:"MISS"})-[:TARGETED]->(server2:server) WHERE id(event1) = $that.data.event1 AND duration("PT45M") > duration.between(localdatetime(event1.timestamp, "yyyy-MM-dd HH:mm:ss.SSSSSS"), localdatetime(event2.timestamp, "yyyy-MM-dd HH:mm:ss.SSSSSS")) > duration("PT5M") AND event1.client_asn = event2.client_asn AND id(server1) = id(server2) AND id(event1) <> id(event2) MATCH (missEvents) WHERE id(missEvents) = idFrom('missEvents', server1.server_id, asset.full_path) SET missEvents.asset = event1.path, missEvents.server = event1.server_id, missEvents.pop = event1.pop, missEvents.firstMiss = coll.min([event1.timestamp, coalesce(missEvents.firstMiss, event1.timestamp)]), missEvents.latestMiss = coll.max([event1.timestamp, coalesce(missEvents.latestMiss, event1.timestamp)]), missEvents: missEvents CREATE (asset)-[:HAD]->(missEvents)-[:FROM]->(server1)<-[:TARGETED]-(event1), (server1)<-[:TARGETED]-(event2) WITH missEvents CALL incrementCounter(missEvents, "cumulativeCount", 1) YIELD count AS cumulativeCount MATCH (missEvents) WHERE missEvents.cumulativeCount = 10 RETURN 'http://localhost:8080/#' + text.urlencode('MATCH(missEvents:missEvents) WHERE id(missEvents)="' + toString(strId(missEvents)) + '" MATCH (event {cache_class:"MISS"})-[:TARGETED]->(server)<-[:FROM]-(missEvents)<-[:HAD]-(asset)<-[:REQUESTED]-(event {cache_class:"MISS"}) RETURN DISTINCT missEvents, event, server, asset LIMIT 10') AS Alert",
"andThen": {
"type": "PrintToStandardOut"
}
}
}
}
]
Once Quine detects the pattern, the event is sent to a standing query output for additional processing and action.
outputs:
cacheMissAlert:
type: Cypher
query: |-
// Add constraints to the cache MISS events match involving the same server and asset pair.
MATCH (server1:server)<-[:TARGETED]-(event1 {cache_class:"MISS"})-[:REQUESTED]->(asset)<-[:REQUESTED]-(event2 {cache_class:"MISS"})-[:TARGETED]->(server2:server)
WHERE id(event1) = $that.data.event1
// Time between consecutive cache MISSes between 5-45 minutes expressed in ISO 8601 duration format (https://en.wikipedia.org/wiki/ISO_8601#Durations)
// Feel free to alter the range to meet your requirements
AND duration("PT45M") > duration.between(localdatetime(event1.timestamp, "yyyy-MM-dd HH:mm:ss.SSSSSS"), localdatetime(event2.timestamp, "yyyy-MM-dd HH:mm:ss.SSSSSS")) > duration("PT5M")
AND event1.client_asn = event2.client_asn
AND id(server1) = id(server2)
AND id(event1) <> id(event2)
////////////////////////////////////////////////////////
// missEvents
////////////////////////////////////////////////////////
// Manifest missEvents node to track metadata relative to consecutive cache MISSes that match the previous constraints
MATCH (missEvents)
WHERE id(missEvents) = idFrom('missEvents', server1.server_id, asset.full_path)
SET missEvents.asset = event1.path,
missEvents.server = event1.server_id,
missEvents.pop = event1.pop,
missEvents.firstMiss = coll.min([event1.timestamp, coalesce(missEvents.firstMiss, event1.timestamp)]),
missEvents.latestMiss = coll.max([event1.timestamp, coalesce(missEvents.latestMiss, event1.timestamp)]),
missEvents: missEvents
// Create subgraph from consecutive cache MISS events to provide a visualization in the Quine Exploration UI
CREATE (asset)-[:HAD]->(missEvents)-[:FROM]->(server1)<-[:TARGETED]-(event1),
(server1)<-[:TARGETED]-(event2)
// Increment the missEvents counter for the purpose of triggering an alert at a specified threshold
WITH missEvents CALL incrementCounter(missEvents, "cumulativeCount", 1) YIELD count AS cumulativeCount
// Trigger alert (RETURN clause) that prints URL to local running Quine instance
MATCH (missEvents)
// Threshold at which to emit alert
// Feel free to alter it to meet your requirements
WHERE missEvents.cumulativeCount = 10
RETURN 'http://localhost:8080/#' + text.urlencode('MATCH(missEvents:missEvents) WHERE id(missEvents)="' + toString(strId(missEvents)) + '" MATCH (event {cache_class:"MISS"})-[:TARGETED]->(server)<-[:FROM]-(missEvents)<-[:HAD]-(asset)<-[:REQUESTED]-(event {cache_class:"MISS"}) RETURN DISTINCT missEvents, event, server, asset LIMIT 10') AS Alert
andThen:
type: PrintToStandardOut
The result once the pattern is detected is to output a link to the console that an analyst can use to review the event further within Quine's Exploration UI.
2023-02-03 16:00:43,345 Standing query `cacheMissAlert` match: {"meta":{"isPositiveMatch":true,"resultId":"0e38c93e-338c-e964-8867-8487eb083e5b"},"data":{"Alert":"http://localhost:8080/#MATCH%28missEvents%3AmissEvents%29%20WHERE%20id%28missEvents%29%3D%2263c2f862-ea0f-3a3a-9f4a-09b11f176ad0%22%20MATCH%20%28event%20%7Bcache_class%3A%22MISS%22%7D%29-%5B%3ATARGETED%5D-%3E%28server%29%3C-%5B%3AFROM%5D-%28missEvents%29%3C-%5B%3AHAD%5D-%28asset%29%3C-%5B%3AREQUESTED%5D-%28event%20%7Bcache_class%3A%22MISS%22%7D%29%20RETURN%20DISTINCT%20missEvents%2C%20event%2C%20server%2C%20asset%20LIMIT%2010"}}
Running the Recipe¶
❯ java -jar quine-1.7.3.jar -r cdn.yaml --recipe-value in_file=cdn_data_50k.json
Graph is ready
Running Recipe: CDN Cache Efficiency By Segment
Using 11 node appearances
Using 14 quick queries
Using 9 sample queries
2023-02-03 17:05:00,342 WARN [NotFromActor] [graph-service-akka.quine.graph-shard-dispatcher-18] com.thatdot.quine.app.StandingQueryResultOutput$ - Could not verify that the provided Cypher query is idempotent. If timeouts or external system errors occur, query execution may be retried and duplicate data may be created. To avoid this, set shouldRetry = false in the Standing Query output
Running Standing Query STANDING-1
2023-02-03 17:05:00,847 WARN [NotFromActor] [graph-service-akka.quine.graph-shard-dispatcher-18] com.thatdot.quine.app.ingest.serialization.CypherJsonInputFormat - Could not verify that the provided ingest query is idempotent. If timeouts occur, query execution may be retried and duplicate data may be created.
Running Ingest Stream INGEST-1
Quine web server available at http://localhost:8080
| => STANDING-1 count 4248
| => INGEST-1 status is running and ingested 5560
Summary¶
When the standing query detects the cache miss pattern, it will output a link to the console that can be copied and pasted into a browser to explore the event in the Quine Exploration UI. Copy and paste the URL section of the match JSON from your console into your browser.
The nodes will be jumbled together when you first open the graph. Arrange the nodes to look similar to the image below before you start exploring.
Tip
Quick Queries are available by right clicking on a node.
Quick Query | Node Type | Description |
---|---|---|
Adjacent Nodes | All | Display the nodes that are adjacent to this node. |
Refresh | All | Refresh the content stored in a node |
Local Properties | All | Display the properties stored by the node |
Reset Counter | missedEvent | Deletes the missedEvent |
Server Pop | server | Displays the associated PoP |
Cache Hit/Miss Percentage | server | Calculates the Hit/Miss percentages for the server |
PoP Hit/Miss Percentage | pop | Calculates the Hit/Miss percentages for the PoP |
PoP Origins | pop | Displays the asset origins the PoP is serving |
Origin Hit/Miss Percentage | origin | Calculates the Hit/Miss percentages for the origin |
Client Hit/Miss Percentage | client | Calculates the Hit/Miss percentages for the client |
clientGeo Hit/Miss Percentage | clientGeo | Calculates the Hit/Miss percentages for the clientGeo |
Asset Hit/Miss Percentage | asset | Calculates the Hit/Miss percentages for the asset |
Client Geo | asn | Displays the clientGeo associated with the ASN |
ASN Hit/Miss Percentage | asn |