320 likes | 551 Views
40248. Diagnostics and Verification of Oracle Real Application Clusters 10 g. Jack Cai Principal Product Manager Oracle Corporation. Agenda. Real Application Clusters 10 g is for Everyone Benefits of RAC Diagnostics and Verification Diagnostics Framework Cluster Verification Framework
E N D
40248 Diagnostics and Verification of Oracle Real Application Clusters 10g Jack Cai Principal Product Manager Oracle Corporation
Agenda • Real Application Clusters 10g is for Everyone • Benefits of RAC Diagnostics and Verification • Diagnostics Framework • Cluster Verification Framework • Q&A
Oracle RAC 10g for Everyone • Foundation of Oracle grid computing • Much easier to use and manage • Full stack reduces complication • Clusterware • Cluster filesystem • Automatic storage manager • Integral to Enterprise Manager • Workload management with policy based service • Cluster diagnostics and verification
Diagnostics & Verification Benefits • Improve problem resolution time • Reduce need to reproduce problems • Eliminate problems related to incorrect cluster configuration • Improve quality of services through a more systematic approach
Diagnostics Framework • Trace instrumentation of Database Server • Clusterization of oradebug, hang analyzer • Trace loader • Navigator • Equally useful on single instance database • For database server internal diagnostics ONLY • Not for diagnosis of application problems/tuning • Use them ONLY under Oracle support supervisions
Diagnostics Framework Goals • Generate enough diagnostic data for first pass analysis • Minimal overhead to database server • Easy to use • Single system view of a cluster • Tool for data gathering, conversion and repository • Tool for display, navigate and analysis
Instance 1 Instance 2 Instance 3 SGA SGA SGA DIAG Process DIAG Process DIAG Process Diagnostics Architecture (Cluster View) On-line Tools Oradebug X$VIEWs Trace Navigation Tool Trace Loader Off-line Tools Trace Files
X$ Views Instance Process Pi SGA On-line Tools Trace Buffer Pi Process Pj Oradebug DIAG Process Trace Buffer Pj Trace Navigational Tool Trace Loader Trace Files Off-line Tools Diagnostics Architecture (Instance)
Flexible Trace Instrumentation • Focus on issues encountered by customers • Instrument trace on events Oracle support handles most frequently • Foundation laid in 9i, more component instrumentation in 10g • Trace control • Trace is event based, events can be anything within a program • Trace is controllable on a per process basis • 1000 event id (10000-10999), 256 opcodes, 256 levels • Trace is generated to SGA buffer rather than directly to files • X$Trace, X$Trace_events can be queried during run time • “Always-on” minimal tracing • Optional trace archiving
Trace Control Example • Dynamic control • ALTER TRACING <cmd> • ON • OFF • ENABLE <event-string> • DISABLE <event-spec> • FLUSH <proc-spec> • Examples • alter tracing enable <event>12345:<level>5:<pid>32 • alter tracing disable 12345:5:BGS • alter tracing flush 32
oradebug Is Cluster Aware • Single system image of the cluster • Commands are clusterized • DIAG based implementation • Damen process • Minimal overhead • HANGANALYZE analyzes the entire cluster • Examples SQL> oradebug -g <inst_list>/ALL <cmd> SQL> oradebug setinst “3 4”
Offline Trace Loader • Offline trace loader converts binary trace files into text format, and loads files into Oracle databases for archiving • Better data analysis through power of database queries • Build a repository for problem tracking • Facilitate trace data exchange in cross-platform environments
Data Conversion Data Loading Trace Loader Trace Loader Raw Trace Files (in binary or text format) Converted Trace Files (in text format) Target Database Offline Trace Loader • Extracts trace data from archived trace files • Converts binary trace data to text • Loads trace data into database as trace repository • Simple command line interface • trcldr [option,...] • trcldr mode=append ifile=f1.trw userid=scott/tiger
Trace Navigation Tool • Navigate traces with links and highlights across cluster • send/recc, malloc/free • User Defined Protocol Navigation • Coloring of traces based upon attributes • Well-defined queries against keywords • Cluster wide time-ordered event diagram
Summary of Diagnostics • Diagnostics facilities allow support to resolve issues more quickly • First pass failure diagnosis • Allows developers to find bugs faster • Tools are not meant for end users • End users should only use the tools under guidance from Oracle support personnel
Cluster Verification Goal • Eliminate errors through: • Pre/post validation of installation steps and/or configuration changes • Ongoing cluster validation • Flexible • API invoked by other programs • OUI, OEM, DBCA, ASM • Command line interface • Can be run anytime • To be released post Oracle Database 10g
Cluster Verification Framework • Verification categories • Stage verification • Component verification • Diagnostic mode • API for vendor verification tools • Utilizing vendor platforms’ verification capabilities
Stage Verification • RAC deployment process is divided into stages • Stages are steps of installation/configuration changes • All RAC deployment stages are verified • Verification performs pre-stage and post-stage checks • Example stages • Cluster filesystem setup • Clusterware setup • Database installation • Cluster node addition • Database configuration
Component Verification • RAC is divided into key components • Each component can be individually verified • Examples • Node connectivity • Shared storage accessibility • Cluster integrity • System requirements • CFS integrity
Diagnostic Mode • Attempt to pinpoint root cause of a verification failure in diagnostic mode • Applicable to stage and component verification • Specific testings are performed when a verification fails • Only triggered by verification failure
How Verification is done • Command line tool: cluvfy • cluvfy stage –post HWOS (verifies hardware and OS stage) • cluvfy comp sys –p RDBMS (verifies system components for RDBMS installation) • APIs by other relevant tools
Summary • Diagnostics speed up problem resolutions • Cluster verification eliminates incorrect configuration • Ultimately reducing unscheduled system down time and improving system availability • Diagnostics tools should ONLY be used under support guidance • End Users can use cluvfy to perform verifications
Next Steps…. • Recommended sessions • 40131: Oracle Real Application Clusters 10g: The Fourth Generation • 40132: Oracle Real Application Clusters 10g: Manage your cluster Without it Managing You • Recommended demos and hands-on labs • Visit DEMOgrounds under Real Application Clusters • See Your Business in Our Software • Visit the DEMOgrounds for more information. • Relevant web sites to visit for more information • http://otn.oracle.com/products/database/clustering/content.html
Reminder – please complete the OracleWorld online session surveyThank you.
Q & Q U E S T I O N S A N S W E R S A