11Source DB Performance Considerations Speed/power of machineReplication Plus – offload read only clients to target databaseAI files – fixed extentsAI/BI blocksizes should be the same16 K is the most efficientNeed to truncate ai and bi
12Target DB Performance Considerations System should not be underpowered by comparison to the source systemNo need for after-imagingRead only clients
13Network Performance WAN vs. LAN The bigger the pipe the better Determine size of pipeWhitepaper available on PDSN
14Sample Network Bandwidth Calculation Hourly After Image Blocks from 5 production databases = 713mbBy calculation:Replication size = 1.5 * AI size713mb*1.5 = mbAdd replication overhead (1.1)1069.5mb*1.1 = mbThroughput per second:mb/3600=.32679mb/bChange to kilobytes /s:.32679mb/s * 1000 = kb/b needed to sustain transfer
15Replication Performance – DBService Buffer DBService Buffer Queue must be sized appropriately – Solution P121969Failure to do so may result in source slowdown and target falling behindAnalyze AI activity prior to implementingUse PROMON -> R&D -> Status Display (#1)-> DB Service Manager (#16).
16Determining Optimal -Pica The larger the –pica, the further behind the target may getMonitor AI writes during busiest period for one hour-pica = (TAIW/Blockcount) * 1.25Blockcount depends on versionOpenEdge 10 – blockcount is 9.16Progress® 9.1x – blockcount is 18.2
17Sample –Pica Calculation Values up to 8192k for 10.1B01 and later34560 TAIW over one hourFormula for OE 10:(34560/9.16) * 1.25 =Why not just use 8192???
18“Houston, we have a problem” - Promon 03/03/ Status: Database Service Manager17:15:02Communication Area Size : KBTotal Message Entries :Free Message Entries :Used Message Entries :Registered Database Service ObjectsName Rdy Status Messages Locked byOpenEdge Replication Server Y RUNOpenEdge RDBMS Y REGOpenEdge DB Agent Y RUN
19Agenda Terminology and Architecture - visual Replication performance hotspotsReplication availability hotspotsSource and target management19
21Availability Hotspots Loss of DatabaseTCP or Process FailurePrimary MachineSecondary MachineR/WriteClientsAI ExtentsDB Service QRepl ServerSource DatabaseDatabase Brk/Server ProcessesDOWN!LockedRepl AgentDatabase Brk/ServerProcessesTarget DatabaseR/OClientsSevered21
22Availability Hotspots – AI Management Replication DOES NOT manage AI filesAI Files must be emptied and backed up10.1A AI archiver became available.Database crashes when no empty AI extents.
23Availability Hotspots – Locked AI Files AI “locked” when AI notes not replicated to the target dbCommon issue – often caused by simple maintenance routines and failure to monitor status of replicationConfigure the rpserver and rpagent to detect and handle outagesAll about the *.repl.properties file!!Configure, test well and sleep easier at night
24Availability – Many Locked AI Files Use dsrutil monitor against source and target db. Check for replication shared memoryIf rpagent is running, restart the replication server on the sourceIf rpagent is not running, restart target database and restart replication server
25“Houston, we have a problem” – part 2 Extent: 1 Extent: 3Status: Busy Status: LockedType: Variable Length Type: Variable LengthPath: C:\wrk101c\repl\source.a1 Path: C:\wrk101c\repl\source.a3Size: Size:Used: Used:Start: Wed May 14 14:09: Start: Wed May 14 14:08:Seqno: 5 Seqno: 3Extent: 2 Extent: 4Status: Locked Status: LockedPath: C:\wrk101c\repl\source.a2 Path: C:\wrk101c\repl\source.a4Size: Size:Used: Used:Start: Wed May 14 13:45: Start: Wed May 14 14:08:Seqno: 2 Seqno: 4
26Availability – All AI Files Locked Source database activity comes to a haltCan you afford to bring the db down?If not - disablesitereplicationIf yesstop databaseadd new ai filesprostrct reorderrestart.
29Source and Target DBs Won’t Synchronize Try restarting a few times.-Ma, -Mn, -n match?Changes to .properties files?If it continues to fail, contact supportSource/target db log filesAI filespmmgr.properties filearguments=-logging 2
30Agenda Terminology and architecture - visual Replication performance hotspotsReplication availability hotspotsSource and target management30
31Managing Replication – DSRUTIL Monitor and Status Not immediately obvious when a failure occursProactively verify replication performance and status
32DSRUtil Monitor DSRUTIL source/target –C monitor Attaches to replication shared memory.Cannot connect to replication shared memory. Status = -1Checks status of Server and AgentServer statusAgent status
33DSRUtil source –C monitor OpenEdge Replication Monitor Page 1Database: C:\wrk101c\repl\sourceDatabase is enabled as OpenEdge Replication: SourceServer is: In Normal ProcessingNumber of configured agents:Delay Interval (current / min / max): / 5 / 500Recovery information:State: No recovery being performed Agents needing recovery:Agents connected:Agents in synchronization:Transition information:Type: Manual
34DSRUtil target –C monitor (page 1) Database: C:\wrk101c\repl\targetDatabase is enabled as OpenEdge Replication: TargetAgent:Name: agent1ID:Host name:State: Normal ProcessingReady: YesCritical: NoMethod: AsynchronousAgent is waiting for: NothingMaximum bytes in TCP/IP message:Server/Agent connection time: Wed May 14 13:48:Delay Interval (current / min / max): / 5 / 500Transition information:Type: ManualThe last block received at: Wed May 14 13:51:Activity information:Blocks received:Blocks processed:
35DSRUtil target –C monitor (page 2) Blocks acknowledged:Notes processed:Transactions started:Transactions ended:Synchronization points:AI Block Information:Source RDBMS Block (Seq / Block): / 2516Last Processed Block (Seq / Block): / 2494Latency Information:Repl Server behind Source DB by: second(s)Current Source Database Transaction:Last Transaction Applied to Target:Target Current as of (Target, Source): Wed May 14 13:51: , Wed May 14 13:51: with delta of 000:00:00
36DSRUTIL Status DSRUTIL source/target –C status Return code indicates current state of replication server or agent.Good for automated scriptsGrep for return codeReplication User Guide provides details
37Managing Replication – OE Management OpenEdge Management 10.1B02/3.1B02+ remotely monitor log files10.1B02 AdminServer running on the remote systemAttach OE_DB_Replication Log File Ruleset to the log file monitorAlert/ thrown when message violating rule set appears.