Mesos 1.0 (MesosCon NA 2016)

Mesos will soon reach the 1.0 milestone. In addition to new features, this would mean a more stable user-facing API and stricter support/release guarantees for operators/framework developers. The aim of this talk is to apprise the operators/framework developers/users about the new API and also discuss the support/compatibility guarantees offered by Mesos going forward.

This talk is a sequel to the last year’s MesosCon Seattle talk on “Mesos HTTP API” and continues from where it left off.

This talk will cover the following specific topics:
- Discuss the newly introduced Operator APl.
- Update on the recent improvements to the Framework API.
- Update on client libraries for the new Framework API.
- Release cadence for Mesos going forward.
- Support/Compatibility guarantees for operators/framework developers e.g., backporting of patches etc.

Transcript

1. 1 . 0 Vi n o d K o n e ( v i n o d @ m e s o s p h e re . i o ) A n a n d M a z u m d a r ( a n a n d @ m e s o s p h e re . i o )
2. M E S O S 1 . 0 • 1.0 • Released on July 27th 2016 • Biggest release ever! • 1.0.1 • Released on Aug 24th 2016 • Please use this one!
3. W H A T ’ S I N 1 . 0 ? New APIs Agent rename Unified Containerizer GPUs Windows Dynamic weights Release policy CNI XFS Fine-grained AuthZ Docker Volume
4. W H Y 1 . 0 ? • Versioned and documented APIs • Formal release and upgrade policy
5. M E S O S A P I S S c h e d u l e r M a s t e r A g e n t E x e c u t o r Scheduler API Executor API Framework API Master API Agent API O p e r a t o r O p e r a t o r Operator API Internal API
6. A P I V 1 G O A L S API versioning Backwards compatible API evolution Allow pure language client libraries Work inside firewalls / containers Well documented
7. P R O T O C O L • HTTP 1.1 - RPC based • Based on Calls, Responses and Events • Client opens connections to the master/agent • One (or more) connection(s) to send Calls • A Call might get a Response, stream of Events, or no response
8. F R A M E W O R K P R O T O C O L • Every call is a HTTP POST request • Content-Type: application/json or application/x-protobuf • Successful SUBSCRIBE call results in a “200 OK” streaming response • Record-IO formatted events • Persistent connection • “Mesos-Stream-Id” header (schedulers only for now) • All successful non-SUBSCRIBE calls result in “202 Accepted” • Must include “Mesos-Stream-Id" header (schedulers only for now)
9. W O R K F L O W S c h e d u l e r M a s t e r SUBSCRIBE
10. POST /api/v1/scheduler HTTP/1.1 { "type" : "SUBSCRIBE", "subscribe" : { "framework_info" : { "user" : "foo", "name" : "Example HTTP Framework" } } } W O R K F L O W SUBSCRIBE S c h e d u l e r M a s t e r SUBSCRIBE
11. W O R K F L O W HTTP/1.1 200 OK Content-Type: application/json Transfer-Encoding: chunked Mesos-Stream-Id: 130ae4e3-6b13-4ef4-baa9-9f2e85c3e9af <event length>{ "type": "SUBSCRIBED", "subscribed": { "framework_id": { "value": "12220-3440-12532-2345" }, "heartbeat_interval_seconds": 15 } }<more events> <more events> SUBSCRIBE Streaming Response M a s t e r S c h e d u l e r
12. W O R K F L O W S c h e d u l e r M a s t e r Offers Event <event-length> { "type" : "OFFERS", "offers" : [ { "offer_id":{"value": "12214-23523-O235235"}, "framework_id":{"value": "12124-235325-32425"}, "agent_id":{"value": "12325-23523-S23523"}, "hostname":"agent.host", "resources":[...], "attributes":[...], "executor_ids":[] } ] } A g e n t
13. W O R K F L O W S c h e d u l e r M a s t e r ACCEPT POST /api/v1/scheduler HTTP/1.1 Mesos-Stream-Id: 130ae4e3-6b13-4ef4-baa9-9f2e85c3e9af { "framework_id" : {"value" : "12220-3440-12532-2345"}, "type" : "ACCEPT", "accept" : { "offer_ids" : [ {"value" : "12220-3440-12532-O12"}, {"value" : "12220-3440-12532-O12"} ], "operations" : [ {"type" : "LAUNCH", "launch" : {...}} ], "filters" : {...} } } A g e n t
14. S c h e d u l e r M a s t e r W O R K F L O W ACCEPT Response HTTP/1.1 202 Accepted A g e n t
15. W O R K F L O W S c h e d u l e r M a s t e r ACCEPT Launch Task A g e n t
16. W O R K F L O W S c h e d u l e r M a s t e r Launch Task E x e c u t o r SUBSCRIBE A g e n t ACCEPT
17. W O R K F L O W S c h e d u l e r M a s t e r E x e c u t o r SUBSCRIBE POST /api/v1/executor HTTP/1.1 { "type": "SUBSCRIBE", "executor_id": { "value": "387aa966-8fc5-4428-a794-5a868a60d3eb" }, "framework_id": { "value": "49154f1b-8cf6-4421-bf13-8bd11dccd1f1" } } A g e n t
18. W O R K F L O W S c h e d u l e r HTTP/1.1 200 OK Content-Type: application/json Transfer-Encoding: chunked <event length> { "type" : "SUBSCRIBED", "subscribed" : {…} } <more events> Streaming Response S c h e d u l e r M a s t e r E x e c u t o r A g e n t E x e c u t o r SUBSCRIBE
19. W O R K F L O W S c h e d u l e r HTTP/1.1 200 OK Content-Type: application/json Transfer-Encoding: chunked <event-length> { "type": "LAUNCH", "launch": (…) } <more events> Launch Event S c h e d u l e r M a s t e r E x e c u t o r A g e n t E x e c u t o r
20. S C H E D U L E R D I S C O N N E C T I O N S & PA R T I T I O N S • Master tracks the persistent subscription connection • Reconnect within framework failover timeout • Periodic HEARTBEAT events sent by master
21. O P E R A T O R V 1 A P I • Similar to Framework API • Some Calls get Responses • Streaming support! • Subscribe to events happening on the cluster e.g., SUBSCRIBED, TASK_ADDED, TASK_UPDATED • SUBSCRIBED includes the snapshot of the state
22. W O R K F L O W O p e r a t o r M a s t e r GET_VERSION POST /api/v1/ HTTP/1.1 Host: masterhost:5050 Content-Type: application/json { “type” : “GET_VERSION” }
23. O p e r a t o r M a s t e r GET_VERSION HTTP/1.1 200 OK { “type” : “GET_VERSION”, “get_version”¹ : { “version_info” : { “version“ “ 1.0.1” } } } W O R K F L O W Response
24. W O R K F L O W O p e r a t o r M a s t e r SUBSCRIBE POST /api/v1/ HTTP/1.1 Host: masterhost:5050 Content-Type: application/json { “type” : “SUBSCRIBE” }
25. W O R K F L O W O p e r a t o r M a s t e r SUBSCRIBE Streaming Response HTTP/1.1 200 OK Content-Type: application/json Transfer-Encoding: chunked <event length>{ "type" : “SUBSCRIBED”, “subscribed" : { “get_state" : { “get_tasks” : {} “get_executors” : {} “get_frameworks” : {} “get_agents” : {} } } }<more events>
26. W O R K F L O W O p e r a t o r M a s t e r TASK_ADDED Event <event-length> { "type": "TASK_ADDED", "task_added": { "task": { "name": "dummy-task", "task_id": {"value": "12344-my-task"}, "agent_id": {"value": “1424-32423-S0”}, … } } }<more events>
27. W O R K F L O W O p e r a t o r M a s t e r TASK_UPDATED Event <event length>{ "type" : “TASK_UPDATED”, “task_updated" : { "task_id" : { "value" : “12344-my-task"}, “framework_id" : { "value" : “1144-12-123"}, "agent_id" : { "value" : "1424-32423-S0"}, "state" : "TASK_RUNNING", } }<more events>
28. A P I S TA T U S • Scheduler API & Executor API : Stable • Operator API : Experimental • Client library implementations • http://mesos.apache.org/documentation/latest/ api-client-libraries/
29. http://mesos.apache.org/documentation/latest/versioning/
30. R E L E A S E P O L I C Y • New release every 2 months • Supported for 6 months • Patch releases every month • Release branches to ease back ports
31. U P G R A D E P O L I C Y • Upgrade from any minor version to any minor version! • 1.2 —> 1.8 (no need to go through 1.3 to 1.7) • Caveat: Doesn’t apply to deprecated features • Deprecation period for a feature is 6 months
32. W H A T ’ S N E X T ? M a s t e r A g e n t HTTP Internal API
33. W H A T ’ S N E X T ? S c h e d u l e r M a s t e r A g e n t E x e c u t o r /api/v1/scheduler /api/v1/executor /api/v1/agent /api/v1 /api/v1 O p e r a t o r O p e r a t o r
34. U N I F I E D A P I S c h e d u l e r M a s t e r A g e n t E x e c u t o r /api/v2 /api/v2 /api/v2 /api/v2 /api/v2 O p e r a t o r O p e r a t o r
35. P R O T O C O L • gRPC • Auto-generated client libs! • HTTP/2 • Single bi-directional connection!
36. T H A N K Y O U http://mesos.apache.org/blog/mesos-1-0-0-released/