Documente Academic
Documente Profesional
Documente Cultură
Version 2.0
Purpose
Before a new computer /cluster system is deployed in production it is important to test the system thoroughly to validate that it will perform at a satisfactory level, relative to its service level objectives. Testing is also required when introducing major or minor changes to the system. This document provides an outline consisting of basic guidelines and recommendations for testing a new RAC system. This test plan outline can be used as a framework for building a system test plan specific to each companys RAC implementation and their associated service level objectives.
Testing Objectives
In addition to application functionality testing, overall system testing is normally performed for one or more of the following reasons: Verify that the system has been installed and configured correctly. Check that nothing is broken. Establish a baseline of functionality behavior such that we can answer the question down the road: has this ever worked in this environment? Verify that basic functionality still works in a specific environment and for a specific workload. Vendors normally test their products very thoroughly, but it is not possible to test all possible hardware/software combinations and unique workloads. Make sure that the system will achieve its objectives, in particular, availability and performance objectives. This can be very complex and normally requires some form of simulated production environment and workload. Test operational procedures. This includes normal operational procedures and recovery procedures. Train operations staff.
Testing should be performed in an environment that mirrors the production environment as much as possible. The software configuration should be identical but for cost reasons it might be necessary to use a scaled down hardware configuration. All testing should be performed while running a workload that is as close to production as possible. When planning for system testing it is extremely important to understand how the application has been designed to handle the failures outlined in this plan and to ensure that the expected results are met at the application level as well as the database level. Oracle technologies that enable fault tolerance of the database at the application level include the following: Fast Application Notification (FAN) Notification mechanism that alerts application of service level changes of the database. Fast Connection Failover (FCF) Utilizes FAN events to enable database clients to proactively react to down events by quickly failing over connections to surviving database instances. Transparent Application Failover (TAF) Allows for connections to be automatically reestablished to a surviving database instance in the case that the instance servicing the initial connection should fail. TAF has the ability to fail over in-flight select statements (if configured) but insert, update and delete transactions will be rolled back. Runtime Connection Load Balancing (RCLB) Provides intelligence about the current service level of the database instances to application connection pools. This increases the performance of the application by utilizing least loaded servers to service application requests and allows for dynamic workload balancing in the event of the loss of service by a database instance or increase of service by adding a database instance. More information on each of the above technologies can be found in the Oracle Real Application Clusters Administration and Deployment Guide 11g Release 2. Generating a realistic application workload can be complex and expensive but it is the most important factor for effective testing. For each individual test in the plan, a clear understanding of the following is required: What is the objective of the test and how does this relate to the overall system objectives? Exactly how will the test be performed and what are the execution steps? What are the success/failure criteria, and what are the expected results? How will the test result be measured? Which tools will be used? Which logfiles and other data will be collected? Which operational procedures are relevant? What are the expected results of the application for each of the defined tests (TAF, FCF, RCLB)?
and/or additional tests might have to be constructed. Introducing multiple failures at the same time should also be considered. This list only covers testing for RAC-related components and procedures. Additional tests are required for other parts of the system. These tests should be performed with a realistic workload on the system. Procedures for detecting and recovering from these failures must also be tested. In some worst-case scenarios it might not be possible to recover the system within an acceptable time frame and a disaster recovery plan should specify how to switch to an alternative system or location. This should also be tested. The result of a test should initially be measured at a business or user level to see if the result is within the service level agreement. If a test fails it will be necessary to gather and analyze the relevant log and trace files. The analysis can result in system tuning, changing the system architecture or possibly reporting component problems to the appropriate vendor. Also, if the system objectives turn out to be unrealistic, they might have to be changed.
Test # Test 2
Procedure Start client workload. Identify the node that is the OCR master using the following grep command from any of the nodes: grep -i "OCR MASTER" $GI_HOME/log/<node_name>/crsd/crsd.l* NOTE: Windows users must manually review the $GI_HOME/log/<node_name>/crsd/crsd.l* logs to determine the OCR Master. Power off the node that is the OCR master. NOTE: On many servers the power-off switch will perform a controlled shutdown, and it might be necessary to cut the power supply
Actual Results/Notes
Test 3
On clusters having 3 or fewer nodes, one of the SCAN VIPs and Listeners will be relocated to the restarted node when the Oracle Clusterware starts. The VIP will migrate back to the restarted node. Services that had failed over as a result of the node failure will NOT automatically be relocated. Failed resources (asm, listener, instance, etc) will be restarted by the Clusterware.
Time for all resources to become available again, Check with crsctl stat res t
Test 4
Issue a reboot on all nodes at the same time o For AIX, HPUX, Windows: shutdown r o For Linux: shutdown r now o For Solaris: reboot
Time for all resources to become available again, Check with crsctl stat res t.
Test # Test 5
Procedure Start client workload Identify single database instance with the most client connections and abnormally terminate that instance: o For AIX, HPUX, Linux, Solaris: Obtain the PID for the pmon process of the database instance: # ps ef | grep pmon kill the pmon process: # kill 9 <pmon pid> o For Windows: Obtain the thread ID of the pmon thread of the database instance by running: SQL> select b.name, p.spid from v$bgprocess b, v$process p where b.paddr=p.addr and b.name=PMON; Run orakill to kill the thread: cmd> orakill <SID> <Thread ID>
Expected Results One of the other instances performs instance recovery Services are moved to available instances, if a preferred instance failed Client connections are moved / reconnected to surviving instances (Procedure and timings will depend on client types and configuration) After a short freeze, surviving instances continue processing the workload Failing instance will be restarted by Oracle Clusterware, unless this feature has been disabled
Test 6
One other instance performs instance recovery Services are moved to available instances, if a preferred instance failed Client connections are moved / reconnected to surviving instances (Procedure and timings will depend on client types and configuration) The instance will NOT be automatically restarted by Oracle Clusterware due to the user invoked shutdown.
Measures Time to detect instance failure Time to complete instance recovery. Check alert log for recovering instance Time to restore client activity to same level (assuming remaining nodes have sufficient capacity to run workload) Duration of database freeze during failover. Time before failed instance is restarted automatically by Oracle Clusterware and is accepting new connections Time to detect instance failure. Time to complete instance recovery. Check alert log for recovering instance. Time to restore client activity to same level (assuming remaining nodes have sufficient capacity to run workload). The instance will NOT be restarted by Oracle Clusterware due to the user induced shutdown.
Actual Results/Notes
Test # Test 7
Procedure Automatic restart by Oracle Clusterware if it is an uncontrolled failure Manual restart necessary if a shutdown command was issued. Manual restart when the "Auto Start" option for the related instance has been disabled.
Test 8
Start client workload Identify a single ASM instance in the cluster: o For AIX, HPUX, Linux, Solaris: Obtain the PID for the pmon process of the ASM instance: # ps ef | grep pmon kill the pmon process: # kill 9 <pmon pid> o For Windows: Obtain the thread ID of the pmon thread of the ASM instance by running: SQL> select b.name, p.spid from v$bgprocess b, v$process p where b.paddr=p.addr and b.name=PMON; Run orakill to kill the thread: cmd> orakill <SID> <Thread ID>
Expected Results Instance rejoins RAC cluster without any problems (review alert logs etc.) Client connections and workload will be load balanced across the new instance (Manual procedure might be required to redistribute workload if long running / permanent connections) The *.dg, *.acfs, *.asm and *.db resources that were running on that node will go offline (crsctl stat res t). By default these resources will be automatically restarted by Oracle Clusterware. One other instance performs instance recovery Services are moved to available instances, if a preferred instance failed Client connections are moved / reconnected to surviving instances (Procedure and timings will depend on client types and configuration) After the database reconfiguration is complete, surviving instances continue processing the workload The Clusterware alert log will show crsd going offline due to an inaccessible OCR if the OCR is stored in ASM. CRSD will automatically restart
Measures Time before services and workload are rebalanced across all instances (including any manual steps)
Actual Results/Notes
Time to detect instance failure Time to complete instance recovery. Check alert log for recovering instance Time to restore client activity to same level (assuming remaining nodes have sufficient capacity to run workload) Duration of database reconfiguration. Time before failed resources are restarted and the database instance is accepting new connections
Test # Test 9
Procedure Start client workload Abnormally terminate 2 different database instances from the same database at the same time: o For AIX, HPUX, Linux, Solaris: Obtain the PID for the pmon process of the database instance: # ps ef | grep pmon kill the pmon process: # kill 9 <pmon pid> o For Windows: Obtain the thread ID of the pmon thread of the database instance by running: SQL> select b.name, p.spid from v$bgprocess b, v$process p where b.paddr=p.addr and b.name=PMON; Run orakill to kill the thread: cmd> orakill <SID> <Thread ID>
Expected Results Same as instance failure. Both instances should be recovered and restarted without problems.
Actual Results/Notes
Test # Test 10
Procedure For AIX, HPUX, Linux and Solaris: Obtain the PID for the listener process: # ps ef | grep tnslsnr Kill the pmon process: # kill 9 <listener pid> For Windows: Use Process Explorer to identify the tnslistener.exe process for the database listener. This will be the tnslistener.exe registered to the <home name>TNSListener service (not the <home name>TNSListenerLISTENER_SCAN<n> service). Once the proper tnslistener.exe is identified kill the process by right clicking the executable and choosing Kill Process.
Expected Results No impact on connected database sessions. New connections are redirected to listener on other node (depends on client configuration) Local database instance will receive new connections if shared server is used. Local database instance will NOT receive new connections if dedicated server is used. The Listener failure is detected by the ORAAGENT and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ crsd/crsd.log o $GI_HOME/log/<nodename>/ agent/crsd/oraagent_<GI_own er>/oraagent_<GI_owner>.log
Measures Time for the Clusterware to detect failure and restart listener.
Actual Results/Notes
Test 11
For AIX, HPUX, Linux and Solaris: Obtain the PID for the SCAN listener process: # ps ef | grep tnslsnr Kill the pmon process: # kill 9 <listener pid> For Windows: Use Process Explorer to identify the tnslistener.exe process for the SCAN listener. This will be the tnslistener.exe registered to the <home name> TNSListenerLISTENER_SCAN<n> service (not the <home name>TNSListener service). Once the proper tnslistener.exe is identified kill the process by right clicking the executable and choosing Kill Process.
No impact on connected database sessions. New connections are redirected to listener on other node (depends on client configuration) The Listener failure is detected by CRSD ORAAGENT and is automatically restarted. Review the following logs: $GI_HOME/log/<nodename>/cr sd/crsd.log $GI_HOME/log/<nodename>/ag ent/crsd/oraagent_<GI_owner>/o raagent_<GI_owner>.log
Test # Test 12
Procedure Unplug all network cables for the public network NOTE: Configurations using NIS must also have implemented NCSD for this test to succeed with the expected results. NOTE: It is recommended NOT to use ifconfig to down the interface, this may lead to the address still being plumbed to the interface resulting in unexpected results.
Test 13
Assuming dual NICs are configured public interface for redundancy (e.g. bonding, teaming, etc). Unplug the network cable from 1 of the NICs. NOTE: It is recommended NOT to use ifconfig to down the interface, this may lead to the address still being plumbed to the interface resulting in unexpected results.
Expected Results Check with crsctl stat res t o The ora.*.network and listener resources will go offline for the node. o SCAN VIPs and SCAN LISTENERs running on the node will fail over to a surviving node. o The VIP for the node will fail over to a surviving node. The database instance will remain up but will be unregistered with the remote listeners. Database services will fail over to one of the other available nodes. If TAF is configured, clients should fail over to an available instance. Network traffic should fail over to other NIC without impacting any of the cluster resources.
Actual Results/Notes
Time to fail over to other NIC card. With bonding /teaming configured this should be less than 100ms.
Test # Test 14
Procedure Unplug all network cables for the interconnect network NOTE: It is recommended NOT to use ifconfig to down the interface, this may lead to the address still being plumbed to the interface resulting in unexpected results.
Note: The method in which a node is evicted has changed in 11.2.0.2 with the introduction of a new feature called Reboot less Restart. Reboot less restart aims to achieve a node eviction without actually rebooting the node.
Expected Results For 11.2.0.1: CSSD will detect split-brain situation and perform one of the following: o In a two-node cluster the node with the lowest node number will survive the other node will be rebooted. o In a multiple node cluster the largest sub-cluster will survive others will be rebooted. Review the following logs: o $GI_HOME/log/<nodename>/ cssd/ocssd.log o $GI_HOME/log/<nodename>/ alert<nodename>.log
Measures Time to detect split brain and start eviction. For 11.2.0.1: See measures for node failure.
Actual Results/Notes
Procedure Unplug all network cables for the interconnect network NOTE: It is recommended NOT to use ifconfig to down the interface, this may lead to the address still being plumbed to the interface resulting in unexpected results.
Note: The method in which a node is evicted has changed in 11.2.0.2 with the introduction of a new feature called Reboot less Restart. Reboot less restart aims to achieve a node eviction without actually rebooting the node.
Expected Results For 11.2.0.2 and above: CSSD will detect split-brain situation and perform one of the following: o In a two-node cluster the node with the lowest node number will survive. o In a multiple node cluster the largest sub-cluster will survive. On the node(s) that is being evicted, a graceful shutdown of Oracle Clusterware will be attempted. o All I/O capable client processes will be terminated and all resources will be cleaned up. If process termination and/or resource cleanup does not complete successfully the node will be rebooted. o Assuming that the above has completed successfully, OHASD will attempt to restart the stack. In this case the stack will be restarted once the network connectivity of the private interconnect network has been restored. Review the following logs: o $GI_HOME/log/<nodename>/ alert<nodename>.log o $GI_HOME/log/<nodename>/ cssd/ocssd.log
Measures For 11.2.0.2 and above: Oracle Clusterware will gracefully shutdown, should graceful shutdown fail (due to I/O processes not being terminated or resource cleanup) the node will be rebooted. Assuming that the graceful shutdown of Oracle Clusterware succeeded, OHASD will restart the stack once network connectivity for the private interconnect has been restored.
Actual Results/Notes
Test # Test 15
Procedure Assuming dual NICs are configured for the private interface for redundancy (e.g. bonding, teaming, etc). Unplug the network cable from 1 of the NICs. NOTE: It is recommended NOT to use ifconfig to down the interface, this may lead to the address still being plumbed to the interface resulting in unexpected results. Assuming 2 or more NICs configured for Oracle Redundant Interconnect and HAIP. Unplug the network cable from 1 of the NICs NOTE: It is recommended NOT to use ifconfig to down the interface, this may lead to the address still being plumbed to the interface resulting in unexpected results.
Expected Results Network traffic should fail over to other NIC without impacting any of the cluster resources.
Measures Time to fail over to other NIC card. With bonding / teaming configured this should be less than 100ms.
Actual Results/Notes
Test 16
Interconnect NIC Failure (Oracle Redundant Interconnect, 11.2.0.2 and higher only) Note: This test is applicable for those on 11.2.0.2 and higher using Oracle Redundant Interconnect/HAIP
Test 17
The HAIP running on the NIC in which the cable was pulled will failover to one of the surviving NICs in the configuration. Clusterware and/or RAC communication will not be impacted. Review the following logs: o $GI_HOME/log/<nodename>/ cssd/ocssd.log o $GI_HOME/log/<nodename>/ gipcd/gipcd.log Upon reconnecting the cable, the HAIP that failed over will relocate back to its original interface. Network traffic should fail over to other switch without any impact on interconnect traffic or instances.
Failover (and fail back) will be seamless (no disruption in service from any node in the cluster).
Time to fail over to other NIC card. With bonding /teaming/11.2 Redundant Interconnect configured this should be less than 100ms.
Test # Test 18
Test Node Loses Access to Disks with CSS Voting Device Note: The method in which a node is evicted has changed in 11.2.0.2 with the introduction of a new feature called Reboot less Restart. Reboot less restart aims to achieve a node eviction without actually rebooting the node.
Procedure Unplug external storage cable connection (SCSI, FC or LAN cable) from one node to disks containing the CSS Voting Device(s). NOTE: To perform this test it may be necessary to isolate the CSS Voting Device(s) to an isolated ASM diskgroup or CFS.
Expected Results For 11.2.0.1: CSS will detect this and evict the node with a reboot. Review the following logs: o $GI_HOME/log/<nodename>/ cssd/ocssd.log o $GI_HOME/log/<nodename>/ alert<nodename>.log For 11.2.0.2 and above: CSS will detect this and evict the node as follows: o All I/O capable client processes will be terminated and all resources will be cleaned up. If process termination and/or resource cleanup does not complete successfully the node will be rebooted. o Assuming that the above has completed successfully, OHASD will attempt to restart the stack. In this case the stack will be restarted once the network connectivity of the private interconnect network has been restored. Review the following logs: o $GI_HOME/log/<nodename>/ alert<nodename>.log o $GI_HOME/log/<nodename>/ cssd/ocssd.log
Measures For 11.2.0.1: See measures for node failure For 11.2.0.2 and above: Oracle Clusterware will gracefully shutdown, should graceful shutdown fail (due to I/O processes not being terminated or resource cleanup) the node will be rebooted. Assuming that the graceful shutdown of Oracle Clusterware succeeded, OHASD will restart the stack once network connectivity for the private interconnect has been restored.
Actual Results/Notes
Test # Test 19
Procedure Unplug external storage cable connection (SCSI, FC or LAN cable) from one node to disks containing the OCR Device(s). NOTE: To perform this test it may be necessary to isolate the OCR Device(s) to an isolated ASM diskgroup or CFS.
Expected Results CRSD will detect the failure of the OCR device and abort. OHASD will attempt to restart CRSD 10 times after which manual intervention will be required. The database instance, ASM instance and listeners will not be impacted. Review the following logs: o $GI_HOME/log/<nodename>/ cssd/crsd.log o $GI_HOME/log/<nodename>/ alert<nodename>.log o $GI_HOME/log/<nodename>/ ohasd/ohasd.log
Measures Monitor database status under load to ensure no service interruption occurs.
Actual Results/Notes
Test 20
Node Loses Access to Single Path of Disk Subsystem (OCR, Voting Device, Database files)
Unplug external storage cable connection (SCSI, FC or LAN cable) from node to disk subsystem.
If multi-pathing is enabled, the multi-pathing configuration should provide failure transparency No impact to database instances.
Test 21
Test 22
Assuming ASM normal redundancy Power off / pull out / offline (depending on config) one ASM disk. Power on / insert / online the ASM disk
No impact on database instances ASM starts rebalancing (view ASM alert logs). No impact on database instances ASM starts rebalancing (view ASM alert logs).
Monitor database status under load to ensure no service interruption occurs. Path failover should be visible in the OS logfiles. Monitor progress: select * from v$asm_operation Monitor progress: select * from v$asm_operation
Test # Test 23
Procedure Remove access to a multiplexed voting disk from all nodes. If voting disks are in a normal redundancy disk group remove access to one of the ASM disks.
Expected Results Cluster will remain available. The voting disk will be automatically brought online when access is restored. Voting Disks can be queried using crsctl query css votedisk. Review the following logs: o $GI_HOME/log/<nodename>/ cssd/cssd.log o $GI_HOME/log/<nodename>/ alert<nodename>.log
Actual Results/Notes
Test 24
1. Remove
access to one copy of OCR or force dismount of ASM diskgroup (asmcmd umount <dg_name> -f). 2. Replace the disk or remount the diskgroup, ocrcheck will report the OCR to be out of sync. 3. Delete the corrupt OCR (ocrconfig delete +<diskgroup>) and read the OCR (ocrconfig add +<diskgroup>). This avoids having to stop CRSD.
There will be no impact on the cluster operation. The loss of access and restoration of the missing/corrupt OCR will be reported in: o $GI_HOME/log/<nodename>/ cssd/crsd.log o $GI_HOME/log/<nodename>/ alert<nodename>.log
There is no impact on the cluster operation The OCR can be replaced online, without a cluster outage.
Test 25
Add a node to the cluster and extend the database (if admin managed) to that node
NOTE: This test assumes that the OCR is mirrored to 2 ASM diskgroups that do not contain voting disks or data or stored on CFS Follow the procedures in Oracle Clusterware Administration and Deployment Guide 11g Release 2 Chapter 4 to extend Grid Infrastructure to the new node. After extending the Grid Infrastructure, follow the procedures in Oracle Real Application Clusters Administration and Deployment Guide 11g Release 2 Chapter 10 to extend the RDBMS binaries and database to the new node.
The new node will successfully be added to the cluster. If the database is policy managed and there is free space in the server pool for the new node the database will be extended to the new node automatically (OMF should be enabled so no user intervention is required). The new database instance will begin servicing connections.
The node is dynamically added to the cluster If the database is policy managed an instance for the database will automatically be created on the new node.
Test # Test 26
Procedure Follow the procedures in Oracle Real Application Clusters Administration and Deployment Guide 11g Release 2 Chapter 10 to delete the node from the cluster. After successfully removing the RDBMS installation, follow the procedures in Oracle Clusterware Administration and Deployment Guide 11g Release 2 Chapter 4 to remove the node from the cluster.
Expected Results The connections on to the database instance being removed will fail over to the remaining instances (if configured). The node will be successfully removed from the cluster.
Actual Results/Notes
Test 2
For Windows: Use Process Explorer to identify the crsd.exe process. Once the crsd.exe process is identified kill the process by right clicking the executable and choosing Kill Process. For AIX, HPUX, Linux and Solaris: Obtain the PID for the EVMD process: # ps ef | grep evmd Kill the EVMD process: # kill 9 <evmd pid>
Expected Results CRSD process failure is detected by the orarootagent and CRSD is restarted. Review the following logs: o $GI_HOME/log/<nodename>/ crsd/crsd.log o $GI_HOME/log/<nodename>/ agent/ohasd/orarootagent_root/ orarootagent_root.log
Actual Results/Notes
For Windows: Use Process Explorer to identify the evmd.exe process. Once the evmd.exe process is identified kill the process by right clicking the executable and choosing Kill Process.
EVMD process failure is detected by the OHASD orarootagent and CRSD is restarted. Review the following logs: o $GI_HOME/log/<nodename>/ evmd/evmd.log o $GI_HOME/log/<nodename>/ agent/ohasd/oraagent_grid /oraagent_grid.log
Test # Test 3
Procedure For AIX, HPUX, Linux and Solaris: Obtain the PID for the CSSD process: # ps ef | grep cssd Kill the CSSD process: # kill 9 <cssd pid>
Expected Results The node will reboot. Cluster reconfiguration will take place
Test 4
CRSD ORAAGENT RDBMS Process Failure NOTE: Test Valid for Only Multi User Installations.
For Windows: Use Process Explorer to identify the ocssd.exe process. Once the ocssd.exe process is identified kill the process by right clicking the executable and choosing Kill Process. For AIX, HPUX, Linux and Solaris: Obtain the PID for the CRSD oraagent for the RDBMS software owner: # cat $GI_HOME/log/<nodename>/agent/crsd/oraag ent_<rdbms_owner>/oraagent_<rdbms_owner> .pid # kill 9 <pid for RDBMS oraagent process>
Measures Time for the eviction and cluster reconfiguration on the surviving nodes Time for the node to come back online and reconfiguration to complete to add the node as an active member of the cluster.
Actual Results/Notes
The ORAAGENT process failure is detected by CRSD and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ crsd/crsd.log o $GI_HOME/log/<nodename>/ agent/crsd/oraagent_<rdbms_o wner>/oraagent_<rdbms_owne r>.log The Grid Infrastructure ORAAGENT process failure is detected by CRSD and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ crsd/crsd.log o $GI_HOME/log/<nodename>/ agent/crsd/oraagent_<GI_own er>/oraagent_<GI_owner>.log
Test 5
For AIX, HPUX, Linux and Solaris: Obtain the PID for the CRSD oraagent for the GI software owner: # cat $GI_HOME/log/<nodename>/agent/crsd/oraag ent_<GI_owner>/oraagent_<GI_owner>.pid # kill 9 <pid for GI oraagent process> For Windows: Use Process Explorer to identify the crsd oraagent.exe process that is a child process of crsd.exe (or obtain the pid for the crsd oraagent.exe as shown in the Unix/Linux instructions above). Once the proper oraagent.exe process is identified kill the process by right clicking the executable and choosing Kill Process.
Test # Test 6
Procedure For AIX, HPUX, Linux and Solaris: Obtain the PID for the CRSD orarootagent: # cat $GI_HOME/log/<nodename>/agent/crsd/oraro otagent_root/orarootagent_root.pid # kill 9 <pid for orarootagent process>
Test 7
For Windows: Use Process Explorer to identify the crsd orarootagent.exe process that is a child process of crsd.exe (or obtain the pid for the crsd orarootagent.exe as shown in the Unix/Linux instructions above). Once the proper orarootagent.exe process is identified kill the process by right clicking the executable and choosing Kill Process. For AIX, HPUX, Linux and Solaris: Obtain the PID for the OHASD oraagent: # cat $GI_HOME/log/<nodename>/agent/ohasd/oraa gent_<GI_owner>/oraagent_<GI_owner>.pid # kill 9 <pid for oraagent process>
Expected Results The ORAROOTAGENT process failure is detected by CRSD and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ crsd/crsd.log o $GI_HOME/log/<nodename>/ agent/crsd/orarootagent_root/o rarootagent_root.log
Actual Results/Notes
For Windows: Use Process Explorer to identify the ohasd oraagent.exe process that is a child process of ohasd.exe (or obtain the pid for the ohasd oraagent.exe as shown in the Unix/Linux instructions above). Once the proper oraagent.exe process is identified kill the process by right clicking the executable and choosing Kill Process.
The ORAAGENT process failure is detected by OHASD and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ ohasd/ohasd.log o $GI_HOME/log/<nodename>/ agent/ohasd/oraagent_<GI_ow ner>/oraagent_<GI_owner>.lo g
Test # Test 8
Procedure For AIX, HPUX, Linux and Solaris: Obtain the PID for the OHASD orarootagent: # cat $GI_HOME/log/<nodename>/agent/ohasd/orar ootagent_root/orarootagent_root.pid # kill 9 <pid for orarootagent process>
Test 9
For Windows: Use Process Explorer to identify the ohasd orarootagent.exe process that is a child process of ohasd.exe (or obtain the pid for the ohasd orarootagent.exe as shown in the Unix/Linux instructions above). Once the proper orarootagent.exe process is identified kill the process by right clicking the executable and choosing Kill Process. For AIX, HPUX, Linux and Solaris: Obtain the PID for the CSSDAGENT: # ps ef | grep cssdagent # kill 9 <pid for cssdagent process>
Expected Results The ORAROOTAGENT process failure is detected by OHASD and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ ohasd/ohasd.log o $GI_HOME/log/<nodename>/ agent/ohasd/orarootagent_root/ orarootagent_root.log
Actual Results/Notes
Test 10
For Windows: Use Process Explorer to identify the cssdagent.exe process. Once the cssdagent.exe process is identified kill the process by right clicking the executable and choosing Kill Process. For AIX, HPUX, Linux and Solaris: Obtain the PID for the CSSDMONITOR: # ps ef | grep cssdmonitor # kill 9 <pid for cssdmonitor process>
The CSSDAGENT process failure is detected by OHASD and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ ohasd/ohasd.log o $GI_HOME/log/<nodename>/ agent/ohasd/oracssdagent_root /oracssdagent_root.log
For Windows: Use Process Explorer to identify the cssdmonitor.exe process. Once the cssdmonitor.exe process is identified kill the process by right clicking the executable and choosing Kill Process.
The CSSDMONITOR process failure is detected by OHASD and is automatically restarted. Review the following logs: o $GI_HOME/log/<nodename>/ ohasd/ohasd.log o $GI_HOME/log/<nodename>/ agent/ohasd/oracssdmonitor_r oot/oracssdmonitor_root.log
Cluster Infrastructure
To simplify testing and problem diagnosis it is often very useful to do some basic testing on the cluster infrastructure without Oracle software or a workload running. Normally this testing will be performed after installing the hardware and operating system, but before installing any Oracle software. If problems are encountered during System Stress Test or Destructive Testing, diagnosis and analysis can be facilitated by testing the cluster infrastructure separately. Typically some of these destructive tests will be used: Node Failure. Obviously without Oracle software or workload. Restart Failed Node Reboot all nodes at the same time Lost disk access HBA failover. Assuming multiple HBAs with failover capability. Disk controller failover. Assuming multiple disk controllers with failover capability. Public NIC Failure Interconnect NIC Failure NAS (Netapps) storage failure In case of a complete mirror failure, measure the time that the storage reconfiguration needed to be completed. Check the same if going into maintenance mode. If using non-Oracle cluster software: Interconnect Network Failure Lost access to cluster voting/quorum disk
Expected Results/Measures The newly added LUN will appear as a candidate disk within ASM.
Actual Results/Notes
Test 2
Create an external redundancy ASM diskgroup using SQL*Plus Create an normal or high redundancy ASM diskgroup using SQL*Plus
Login to ASM via SQL*Plus and run: create diskgroup <dg name> external redundancy disk <candidate path> ; Login to ASM via SQL*Plus and run: create diskgroup <dg name> norma lredundancy disk '<candidate1 path>, '<candidate 2 path> ;
A successfully created diskgroup. This diskgroup should also be listed in v$asm_diskgroup. The diskgroup will be registered as a Clusterware resource (crsctl stat res t)
Test 3
A successfully created diskgroup with normal redundancy and two failure groups. For high redundancy, it will create three fail groups. The diskgroup will be registered as a Clulsterware resource (crsctl stat res t)
Test 4
Login to ASM via SQL*Plus and run: alter diskgroup <dg name> add disk '<candidate1 path> ; NOTE: Progress can be monitored by querying v$asm_operation
The disk will be added to the diskgroup and the data will be rebalanced evenly across all disks in the diskgroup.
Test 5
Login to ASM via SQL*Plus and run: alter diskgroup <dg name> drop disk <disk name>; NOTE: Progress can be monitored by querying v$asm_operation
The data from the removed disk will be rebalanced across the remaining disks in the diskgroup. Once the rebalance is complete the disk will have a header_status of FORMER (v$asm_disk) and will be a candidate to be added to another diskgroup.
Test # Test 6
Test Undrop a ASM disk that is currently being dropped using SQL*Plus
Procedure Login to ASM via SQL*Plus and run: alter diskgroup <dg name> drop disk <disk name>; Before the rebalance completes run the following command via SQL*Plus: alter diskgroup <dg name> undrop disk <disk name>; NOTE: Progress can be monitored by querying v$asm_operation
Expected Results/Measures The undrop operation will rollback the drop operation (assuming it has not completed). The disk entry will remain in v$asm_disk as a MEMBER.
Actual Results/Notes
Test 7
Drop a ASM diskgroup using SQL*Plus Modify rebalance power of an active operation using SQL*Plus
Login to ASM via SQL*Plus and run: drop diskgroup <dg name>;
Test 8
Login to ASM via SQL*Plus and run: alter diskgroup <dg name> add disk '<candidate1 path> ; Before the rebalance completes run the following command via SQL*Plus: alter diskgroup <dg name> rebalance power <1 11>;. 1 is the default rebalance power. NOTE: Progress can be monitored by querying v$asm_operation Start all the database instances and query the v$asm_client view in the ASM instances.
The diskgroup will be successfully dropped. The diskgroup will be unregistered as a Clusterware resource (crsctl stat res t) The rebalance power of the current operation will be increased to the specified value. This is visible in the v$asm_operation view.
Test 9
Verify CSS-database communication and ASM files access. Check the internal consistency of disk group metadata using SQL*Plus
Test 10
Login to ASM via SQL*Plus and run: alter diskgroup <name> check all
If there are no internal inconsistencies, the statement Diskgroup altered will be returned (asmcmd will return back to the asmcmd prompt). If inconsistencies are discovered, then appropriate messages are displayed describing the problem.
Expected Results/Measures The newly added LUN will appear as a candidate disk within ASM.
Actual Results/Notes
Test 2
Identify the candidate disks for the diskgroup by running: lsdsk candidate Create a XML config file to define the diskgroup e.g.
<dg name="<dg name>" redundancy="external"> <dsk string="<disk path>" /> <a name="compatible.asm" value="11.1"/> <a name="compatible.rdbms" value="11.1"/> </dg>
A successfully created diskgroup. This diskgroup can be viewed using the lsdg ASMCMD command. The diskgroup will be registered as a Clusterware resource (crsctl stat res t)
Test # Test 3
Procedure Identify the candidate disks for the diskgroup by running: lsdsk candidate Create a XML config file to define the diskgroup e.g.
<dg name="<dg_name>" redundancy="normal"> <fg name="fg1"> <dsk string="<disk path>" /> </fg> <fg name="fg2"> <dsk string="<disk path>" /> </fg> <a name="compatible.asm" value="11.1"/> <a name="compatible.rdbms" value="11.1"/> </dg>
Expected Results/Measures A successfully created diskgroup. This diskgroup can be viewed using the lsdg ASMCMD command. The diskgroup will be registered as a Clusterware resource (crsctl stat res t)
Actual Results/Notes
Test 4
Identify the candidate disk to be added by running: lsdsk candidate Create a XML config file to define the diskgroup change e.g.
<chdg name="<dg name>"> <add> <dsk string="<disk path>"/> </add> </chdg>
The disk will be added to the diskgroup and the data will be rebalanced evenly across all disks in the diskgroup. Progress of the rebalance can be monitored by running the lsop ASMCMD command.
Login to ASM via ASMCMD and run: chdg <config file>.xml NOTE: Progress can be monitored by running lsop
Test # Test 5
Procedure Identify the ASM name for the disk to be dropped from the given diskgroup: lsdsk -G <dg name> -k Create a XML config file to define the diskgroup change e.g.
<chdg name="<dg name>"> <add> <dsk name="<disk name>"/> </add> </chdg>
Expected Results/Measures The data from the removed disk will be rebalanced across the remaining disks in the diskgroup. Once the rebalance is complete the disk will be listed as a candidate (lsdsk candidate) to be added to another diskgroup. Progress can be monitored by running lsop The diskgroup will be unregistered as a Clusterware resource (crsctl stat res t)
Actual Results/Notes
Login to ASM via ASMCMD and run: chdg <config file>.xml NOTE: Progress can be monitored by running lsop
Test 6
Add a disk to a diskgroup (as shown above). Identify the rebalance operation by running lsop via ASMCMD. Before the rebalance completes run the following command via ASMCMD: rebal power <1-11> <dg name>. NOTE: Progress can be monitored by running lsop Login to ASM via ASMCMD and run: dropdg <dg name>;
The rebalance power of the current operation will be increased to the specified value. This is visible with the lsop command.
Test 7
The diskgroup will be successfully dropped. The diskgroup will be unregistered as a Clusterware resource (crsctl stat res t)
Expected Results/Measures The ASM template will be successfully created and visible within the v$asm_template view.
Actual Results/Notes
Test 2
Use the template above and apply it to a new tablespace to be created on the database Login to ASM via SQL*Plus and run: create tablespace test datafile '+<dg name>/my_files(unreliable)' size 10M;
Test 3
Login to ASM via SQL*Plus and run: alter diskgroup <dg name> drop template unreliable; Login to ASM via SQL*Plus and run: alter diskgroup <dg name> add directory '+<dg name>/my_files';
Test 4
You can use the asmcmd tool to check that the new directory name was created in the desired diskgroup.
Test 5
Login to ASM via SQL*Plus and run: alter diskgroup DATA add alias '+DATA/my_files/datafile_alias' for '+<dg name>/ <db name>/DATAFILE/<file name>'; Login to ASM via SQL*Plus and run: alter diskgroup DATA drop alias '+<dg name>/my_files/ datafile_alias ';
The created directory will have an entry in v$asm_directory Verify that the alias exists in v$asm_alias
Test 6
Test # Test 7
Procedure Identify a data file from a running database. Login to ASM via SQL*Plus and run: alter diskgroup data drop file '+<dg name>/<db name>/DATAFILE/<file name>';
Expected Results/Measures This will fail with the following message: ERROR at line 1: ORA-15032: not all alterations performed ORA-15028: ASM file '+DATA/V102/DATAFILE/TEST.269.654602409' not dropped; currently being accessed
Actual Results/Notes
Test 8
Identify a datafile that is no longer used by a database Login to ASM via SQL*Plus and run: alter diskgroup data drop file '+<dg name>/<db name>/DATAFILE/<file name>';
Test # Test 2
Test 3
Procedure Within ASMCMD issue the volinfo a command and take note of the Volume Device path. As the root user create an ACFS filesystem on the ASM Volume as follows: /sbin/mkfs t acfs <volume device path> As the root user execute the following to mount the ACFS filesystem: /sbin/mount t acfs <volume device path> <mount point> NOTE: If acfsutil was not used to register the file system, the dynamic volume must be enabled on the remote nodes before mounting (within ASMCMD run volenable). Use acfsutil to register the ACFS filesystem: /sbin/acfsutil registry a <volume device path> <mount point>
Expected Results/Measures The filesystem will be successfully created. The filesystem attributes can be viewed by running /sbin/acfsutil info fs
Actual Results/Notes
Test 4
The filesystem will be registered with the ACFS registry. This can be validated by running /sbin/acfsutil registry l The filesystem will be automounted on all nodes in the cluster on reboot
Test 5
Perform the following: echo Testing ACFS > <mount point>/testfile Perform a cat command on the file on all nodes in the cluster.
The file will exist on all nodes with the specified contents.
Test 6
Use acfsutil to register the ACFS filesystem: /sbin/acfsutil registry d <volume device path>
The filesystem will be unregistered with the ACFS registry. This can be validated by running /sbin/acfsutil registry l The filesystem will NOT be automounted on all nodes in the cluster on reboot RAC Starter Kit System Test Plan Outline Page 30
Test # Test 7
Test Add an ACFS filesystem as a Clusterware resource NOTE: This is required when using ACFS for a shared RDBMS Home. When ACFS is registered as a CRS resource it should NOT be registered in the ACFS mount registry. Increase the size of a ACFS filesystem
Procedure Execute the following command as root to add a ACFS filesystem as a Clusterware resource: svrctl add filesystem d < volume device path> -v <volume name> -g <dg name> -m <mount point> -u root Start the ACFS filesystem resource: svrctl start filesystem d <volume device path>
Expected Results/Measures The filesystem will be registered as a resource within the Clusterware. This can be validated by running crsctl stat res t The filesystem will be automounted on all nodes in the cluster on reboot
Actual Results/Notes
Test 8
Test 9
Test 10
Add a disk to the diskgroup housing the ACFS filesystem (if necessary) Use acfsutil as the root user to resize the ACFS filesystem: acfsutil size <size><K|M|G> <mount point> Create an ACFS filesystem that is a minimum of 6GB in size. Add the ACFS filesystem as a Clusterware resource. Install the 11gR2 RDBMS on the shared ACFS filesystem (see install guide) Use acfsutil to create a snapshot of an ACFS filesystem: /sbin/acfsutil snap <name> <ACFS mount point> Use acfsutil to delete a previously created snapshot of an ACFS filesystem: /sbin/acfsutil snap delete <name> <ACFS mount point>
The dynamic volume and filesystem will be resized without an outage of the filesystem provided enough free space exists in the diskgroup. Validate with df h.
A snapshot of the ACFS file system will be created under <ACFS mount point>/.ACFS/snaps.
Test 11
The specified snapshot will be deleted and will no longer appear under <ACFS mount point>/.ACFS/snaps.
Test # Test 12
Procedure Dismount the ACFS filesystem to be checked on ALL nodes: o If the filesystem is registered as a Clusterware resource issue srvctl stop filesystem d <device path> to dismount the filesystem on all nodes o If the filesystem is only in the ACFS mount registry or is not registered with Clusterware in any way dismount the filesystem using umount <mount point>. Execute fsck on the ACFS filesystem as follows: sbin/fsck -a -v -y -t acfs <device path> This command will automatically fix any errors (-a), answer yes to any prompts (-y) and provide verbose output (-v).
Expected Results/Measures FSCK will check the specified ACFS filesystem for errors, automatically fix any errors (-a), answer yes to any prompts (-y) and provide verbose output (-v).
Actual Results/Notes
Test # Test 13
Test 14
Procedure Dismount the ACFS filesystem to be deleted on ALL nodes: o If the filesystem is registered as a Clusterware resource issue srvctl stop filesystem d <device path> to dismount the filesystem on all nodes o If the filesystem is only in the ACFS mount registry or is not registered with CRS in any way dismount the filesystem using umount <mount point>. If the filesystem is registered with the ACFS mount registry deregister the mount point using acfsutil as follows: /sbin/acfsutil registry d <device path> Remove the filesystem from the Dynamic Volume using acfsutil: /sbin/acfsutil rmfs <device path> Use ASMCMD to delete a ASM Dynamic Volume: voldelete G <dg name> <vol name>
Expected Results/Measures The ACFS filesystem will be removed from the ASM Dynamic Volume. Attempts to mount the filesystem should now fail.
Actual Results/Notes
The removed Dynamic Volume will no longer be listed in the output of volinfo a. The disk space utilized by the Dynamic Volume will be returned to the diskgroup.
Actual Results/Notes
Test 2
Use dbms_file_transfer.put_file and get_file functions to copy database files (datafiles, archives, etc) into and out of ASM. NOTE: This requires that a database directory be pre-created and available for the source and destination directories. See PL/SQL Guide for dbms_file_transfer details
The put_file and get file functions will copy files successfully to/from filesystem. This provides an alternate option for migrating to ASM, or to simply copy files out of ASM.
Expected Results/Measures
Actual Results/Notes
Add a Disk/LUN to the RAC nodes and configure the Disk/LUN for use by OCFS2. Create the appropriate partition table on the disk and use partprobe to rescan the partition tables. Create the OCFS2 filesystem by running: /sbin/mkfs t ocfs2 <device path> Add the filesystem to /etc/fstab on all nodes Mount the filesystem on all nodes
The OCFS2 filesystem will be created. The OCFS2 filesystem will be mounted on all nodes
Test # Test 2
Procedure
Expected Results/Measures
Actual Results/Notes
Perform the following: echo Testing OCFS2 > <mount point>/testfile Perform a cat command on the file on all nodes in the cluster.
The file will exist on all nodes with the specified contents.
Test 3
Test 4
Verify that the OCFS2 filesystem is available after a system reboot Enable database archive logs to OCFS2 NOTE: If using the OCFS2 filesystem for database files it must be mounted with the following options: rw,datavolume,nointr Create an RMAN on a OCFS2 filesystem NOTE: If using the OCFS2 filesystem for database files it must be mounted with the following options: rw,datavolume,nointr Create a datapump export on a OCFS2 filesystem Validate OCFS2 functionality during node failures.
The OCFS2 filesystem will automatically mount and be accessible to all nodes after a reboot. Archivelog files are created, and available to all nodes on the specified OCFS2 filesystem.
Test 5
Back up ASM based datafiles to OCFS2 filesystem. Execute baseline recovery scenarios (full, point-in-time, datafile).
RMAN backupsets are created, and available to all nodes on the specified OCFS2 filesystem. Recovery scenarios completed with no errors.
Test 6
Using datapump, take an export of the database to an OCFS2 filesystem. Issue a shutdown r now from a single node in the cluster
A full system export should be created without errors or warnings. OCFS2 filesystem should remain available to surviving nodes.
Test 7
Test # Test 8
Test Validate OCFS2 functionality during disk/disk subsystem path failures NOTE: Only applicable on multipath storage environments. Perform a FSCK of a OCFS2 filesystem
Procedure
Expected Results/Measures
Actual Results/Notes
Unplug external storage cable connection (SCSI, FC or LAN cable) from node to disk subsystem.
If multi-pathing is enabled, the multipathing configuration should provide failure transparency No impact to the OCFS2 filesystem. Path failover should be visible in the OS logfiles.
Test 9
Test 10
Dismount the OCFS2 filesystem to be checked on ALL nodes Execute fsck on the OCFS2 filesystem as follows: sbin/fsck -v -y -t ocfs2 <device path> This command will automatically, answer yes to any prompts (-y) and provide verbose output (-v). Check the OCFS2 cluster status on all nodes by issuing /etc/init.d/o2cb status.
FSCK will check the specified OCFS2 filesystem for errors, answer yes to any prompts (-y) and provide verbose output (-v).
Expected Results
Actual Results/Notes
Test 2
Add a Disk/LUN to the RAC nodes and configure the Disk/LUN for use by OCFS. Create the appropriate partition table on the disk and validate disk and partition table is visible on ALL nodes (this can be achieved via diskpart). Assign a drive letter to the logical drive Create the OCFS filesystem by running: cmd> %GI_HOME%\cfs\ocfsformat /m <drive_letter> /c <cluster size> /v <volume name> /f /a Perform the following: Use notepad to create a text file containing the text TESTING OCFS on an OCFS drive. Use notepad to validate that the file exists on all nodes. Issue a reboot
The OCFS filesystem will be created. The OCFS filesystem will be mounted on all nodes
The file will exist on all nodes with the specified contents.
Test 3
Test 4
Verify that the OCFS filesystem is available after a system reboot Enable database archive logs to OCFS
The OCFS filesystem will automatically mount and be accessible to all nodes after a reboot. Archivelog files are created, and available to all nodes on the specified OCFS filesystem.
Test# Test 5
Procedure
Expected Results
Actual Results/Notes
Back up ASM based datafiles to OCFS filesystem. Execute baseline recovery scenarios (full, point-in-time, datafile).
RMAN backupsets are created, and available to all nodes on the specified OCFS filesystem. Recovery scenarios completed with no errors.
Test 6
Test 7
Test 8
Test 9
Create a datapump export on an OCFS filesystem Validate OCFS functionality during node failures. Remove a drive letter and ensure that the letter is reestablished for that partition Run ocfscollect tool
Using datapump, take an export of the database to an OCFS filesystem. Issue a reboot from a single node in the cluster Using Windows disk management use the Change Drive Letter and Paths option to remove a drive letter associated with an OCFS partition. OCFSCollect is available as an attachment to Note: 332872.1
A full system export should be created without errors or warnings. OCFS filesystem should remain available to surviving nodes. OracleClusterVolumeService should restore the drive letter assignment within a short period of time.
A .zap file (rename to .zip and extract). Can be used as a baseline regarding the health of the available OCFS drives.