Pip Detecting the Unexpected in Distributed Systems Janet Wiener Jeff Mogul Mehul Shah Chip Killian Amin Vahdat Patrick Reynolds

page 2 Pip - November 2005 Motivation Distributed systems exhibit complex behaviors Some behaviors are unexpected – Structural bugs Placement or timing of processing and communication – Performance problems Throughput bottlenecks Over- or under-consumption of resources Unexpected interdependencies Parallel, inter-node behavior is hard to capture with serial, single-node tools – Not captured by traditional debuggers, profilers – Not captured by unstructured log files

page 3 Pip - November 2005 Motivation Three target audiences: Primary programmer – Debugging or optimizing his/her own system Secondary programmer – Inheriting a project or joining a programming team – Learning how the system behaves Operator – Monitoring running system for unexpected behavior – Performing regression tests after a change

page 4 Pip - November 2005 Motivation Programmers wish to examine and check system- wide behaviors – Causal paths – Components of end-to-end delay – Attribution of resource consumption Unexpected behavior might indicate a bug Web server App server Database 500ms 2000 page faults

page 5 Pip - November 2005 Pip overview Pip: 1. Captures events from a running system 2. Reconstructs behavior from events 3. Checks behavior against expectations 4. Displays unexpected behavior Both structure and resource violations Goal: help programmers locate and explain bugs Behavior model Application Expectations Pip checker Unexpected structure Resource violations Pip explorer: visualization GUI

page 6 Pip - November 2005 Outline Expressing expected behavior Building a model of actual behavior Exploring application behavior Results – FAB – RanSub – SplitStream

page 7 Pip - November 2005 Describing application behavior Application behavior consists of paths – All events, on any node, related to one high-level operation – Definition of a path is programmer defined – Path is often causal, related to a user request WWW App server DB Parse HTTP Query Send response Run application time

page 8 Pip - November 2005 Describing application behavior Within paths are tasks, messages, and notices – Tasks: processing with start and end points – Messages: send and receive events for any communication Includes network, synchronization (lock/unlock), and timers – Notices: time-stamped strings; essentially log entries WWW App server DB Parse HTTP Query Send response Run application time “Request = /cgi/…”“2096 bytes in response” “done with request 12”

page 9 Pip - November 2005 Expectations: Recognizers Application behavior consists of paths Each recognizer matches paths – A path can match more than one recognizer A recognizer can be a validator, an invalidator, or neither Any path matching zero validators or at least one invalidator is unexpected behavior: bug? validator CGIRequest task(“Parse HTTP”) limit(CPU_TIME, 100ms); notice(m/Request URL:.*/); send(AppServer); recv(AppServer); invalidator DatabaseError notice(m/Database error:.*/);

page 10 Pip - November 2005 Expectations: Recognizers language repeat: matches a ≤ n ≤ b copies of a block xor: matches any one of several blocks call: include another recognizer (macro) future: block matches now or later – done: force named block to match repeat between 1 and 3 { … } xor { branch: … } future F1 { … } … done(F1);

page 11 Pip - November 2005 Expectations: Aggregate expectations Recognizers categorize paths into sets Aggregates make assertions about sets of paths – Count, unique count, resource constraints – Simple math and set operators assert(instances(CGIRequest) > 4); assert(max(CPU_TIME, CGIRequest) < 500ms); assert(max(REAL_TIME, CGIRequest) <= 3*avg(REAL_TIME, CGIRequest));

page 12 Pip - November 2005 Outline Expressing expected behavior Building a model of actual behavior Exploring application behavior Results

page 13 Pip - November 2005 Building a behavior model Sources of events: Annotations in source code – Programmer inserts statements manually Annotations in middleware – Middleware inserts annotations automatically – Faster and less error-prone Passive tracing or interposition – Easier, but less information Or any combination of the above Model consists of paths constructed from events recorded by the running application

page 14 Pip - November 2005 Annotations Set path ID Start/end task Send/receive message Notice WWW App server DB Parse HTTP Query Send response Run application time “Request = /cgi/…”“2096 bytes in response” “done with request 12”

page 15 Pip - November 2005 Automating expectations and annotations Expectations can be generated from behavior model – Create a recognizer for each actual path – Eliminate repetition – Strike a balance between over- and under-specification Annotations can be generated by middleware Automatic annotations in Mace, Sandstorm, J2EE, FAB – Several of our test systems use Mace annotations Behavior model Application Expectations Pip checker Annotations Unexpected behavior

page 16 Pip - November 2005 Checking expectations Traces Categorized paths Reconciliation Events database Paths Path construction Expectation checking Application For each path P For each recognizer R Does R match P? Check each aggregate Expectations Match start/end task, send/receive message Organize events into causal paths

page 17 Pip - November 2005 Exploring behavior Expectations checker generates lists of valid and invalid paths Explore both sets – Why did invalid paths occur? – Is any unexpected behavior misclassified as valid? Insufficiently constrained expectations Pip may be unable to express all expectations Two ways to explore behavior – SQL queries over tables Paths, threads, tasks, messages, notices – Visualization

page 18 Pip - November 2005 Timing and resource properties for one task Causal view of path Visualization: causal paths Caused tasks, messages, and notices on that thread

page 19 Pip - November 2005 Visualization: communication graph Graph view of all host-to-host network traffic

page 20 Pip - November 2005 Visualization: performance graphs Plot per-task or per-path resource metrics – Cumulative distribution (CDF), probability density (PDF), or vs. time Click on a point to see its value and the task/path represented Time (s) Delay (ms)

page 21 Pip - November 2005 Pip vs. printf Both record interesting events to check off-line – Pip imposes structure and automates checking – Generalizes ad hoc approaches Pipprintf Nesting, causal orderUnstructured Time, path, and threadNo context CPU and I/O dataNo resource information Automatic verification using declarative language Verification with ad hoc grep or expect scripts SQL queries“Queries” using Perl scripts Automatic generation for some middleware Manual placement

page 22 Pip - November 2005 Results We have applied Pip to several distributed systems: – FAB: distributed block store – SplitStream: DHT-based multicast protocol – RanSub: tree-based protocol used to build higher-level systems – Others: Bullet, SWORD, Oracle of Bacon We have found unexpected behavior in each system We have fixed bugs in some systems … and used Pip to verify that the behavior was fixed

page 23 Pip - November 2005 Results: SplitStream (DHT-based multicast protocol) 13 bugs found, 12 fixed – 11 found using expectations, 2 found using GUI Structural bug: some nodes have up to 25 children when they should have at most 18 – This bug was fixed and later reoccurred – Root cause #1: variable shadowing – Root cause #2: failed to register a callback How discovered: first in the explorer GUI, confirmed with automated checking

page 24 Pip - November 2005 Results: FAB (distributed block store) 1 bug (so far), fixed – Four protocols checked: read, write, Paxos, membership Performance bug: nodes seeking quorum call self and peers in arbitrary order – Should call self last, to overlap computation – For cached blocks, should call self second-to-last

page 25 Pip - November 2005 Results: RanSub (tree-based protocol) 2 bugs found, 1 fixed Structural bug: during first round of communication, parent nodes send summary messages before hearing from all children – Root cause: uninitialized state variables Performance bug: linear increase in end-to-end delay for the first ~2 minutes – Suspected root cause: data structure listing all discovered nodes

page 26 Pip - November 2005 Future work Further automation of annotations, tracing – Explore tradeoffs between black-box, annotated behavior models Extensible annotations – Application-specific schema for notices Composable expectations for large systems

page 27 Pip - November 2005 Related work Expectations-based systems – PSpec [Perl, 1993] – Meta-level compilation [Engler, 2000] – Paradyn [Miller, 1995] Causal paths – Pinpoint [Chen, 2002] – Magpie [Barham, 2004] – Project5 [Aguilera, 2003] Model checking – MaceMC [Killian, 2006] – VeriSoft [Godefroid, 2005]

page 28 Pip - November 2005 Conclusions Finding unexpected behavior can help us find bugs – Both structure and performance bugs Expectations serve as a high-level external specification – Summary of inter-component behavior and timing – Regression test for structure and performance Some bugs not exposed by expectations can be found through exploring: queries and visualization

Extra slides

page 30 Pip - November 2005 Resource metrics Real time User time, system time – CPU time = user + system – Busy time = CPU time / real time Major and minor page faults (paging and allocation) Voluntary and involuntary context switches Message size and latency Number of messages sent Causal depth of path Number of threads, hosts in path