SuperCluster - Exavm Zones should be built with default scheduling class of TS instead of FSS

Asset ID:	1-72-1618396.1
Update Date:	2015-03-23
Keywords:

Solution Type Problem Resolution Sure

Solution 1618396.1 : SuperCluster - Exavm Zones should be built with default scheduling class of TS instead of FSS

Applies to:

SPARC SuperCluster T4-4 Half Rack - Version All Versions to All Versions [Release All Releases]
SPARC SuperCluster T4-4 - Version All Versions to All Versions [Release All Releases]
SPARC SuperCluster T4-4 Full Rack - Version All Versions to All Versions [Release All Releases]
Oracle SuperCluster T5-8 Hardware - Version All Versions to All Versions [Release All Releases]
Oracle SuperCluster T5-8 Half Rack - Version All Versions to All Versions [Release All Releases]
Oracle Solaris on SPARC (64-bit)
SuperCluster, exavm zones, FSS, TS, Solaris 11, Critical Threads, SSCTUNER, Zone Performance

Symptoms

A review of an explorer output or simple ps -efcZ command shows that both global and non global zone process are running predominantly with FSS scheduling class. Other symptoms may include:

Database and/or ASM LMS and LGWR process are not running with the SuperCluster recommended critical threads scheduling class of FX-60. For more information on this refer to item DB_2 in <Document 1452277.1>
Intermittant and not easily explained performance issues from database residing in the zones.

Changes

SuperCluster LDom in question was configured with exavm zones

Cause

The initial release of exavm zone creation tools set the default pool and init default scheduling class to FSS. This has been corrected via <Defect 18003577>. This defect correction will only correct newly deployed zones and not existing ones in this condition.

Solution

The solution to this problem is in two phases depending on the condition of your system. The first phase will adjust the scheduling class and the next pahse will detail how you can ensure you have the proper version off ssctuner installed to adjust thread scheduling class inside the zones. These will have to be done during a maintenance window because after completion of phase 1 and 2 (or just phase 1 or just phase 2 if that is all your system requires) a reboot of the LDom will be required. This will also reboot all non global zones in the LDom.

Note: The alternative to bringing ssctuner up out of band would be to patch the system to the Jan 2014 Quarterly Full Stack Download Patch. See the SuperCluster Supported Versions document for your hardware type. They are linked at the bottom of this document.

Please also review SuperCluster - ssctuner is not adjusting the scheduling class of all of lms , lgwr and vktm processes to FX-60 <Document 1628298.1>. This is an additional step that has to be done under the supervision of an Oracle badged employee.

The following needs to be performed in the Global Zone only

Phase 1 as root user

poolcfg -c 'modify pool pool_default (string pool.scheduler="TS")'

pooladm –c

dispadmin -d TS

Phase 2 as root user

Please note this phase is not required if you system is patched to the JAN 2014 QFSDP following the patch readme files. A verification of this can be done by running

pkg list entire

pkg list -v ssctuner

and ensuring your entire version string is 0.5.11-0.175.1.11.0.4.0 or above and ssctuner version string is 0.5.11-1.5.9.237 or above.

Download <Patch 18205769> ORACLE SOLARIS EXA-FAMILY 20140131 REPO ISO IMAGE (SPARC/X86 (64-BIT)) from MOS and copy the zip file into each LDom containing exavm zones. Extract the ISO contained in the archive.

# mkdir -p /tmpisomnt

--The build number will be part of the iso name when you extract it

# mount -F hsfs /SoftwareSource/sol-exafamily-<build number>-sc-repo-full.iso /tmpisomnt

# pkg set-publisher -g file:/// tmpisomnt /repo/ exa-family

# /opt/oracle.supercluster/ssctuner/uninstall.ksh

# pkg uninstall ssctuner( yes to rollback changes)

# pkg install ssctuner

# pkg set-publisher -G file:/// tmpisomnt /repo/ exa-family

# umount /tmpisomnt

# rm –fr /tmpisomnt

-- Make sure the service is online

# svcs -a |grep ssctuner

-- Wait at least 10 minutes

# reboot

When the LDom and all zones and corresponding Grid Infrastructure and Databases are up and running verify vktm, lms and lgwr processes in all zones are running at FX-60.

ps -efcZ |egrep 'lms|lgwr|vktm'

References

<NOTE:1567979.1> - Oracle SuperCluster Supported Software Versions - All Hardware Types
<NOTE:1452277.1> - SuperCluster Critical Issues
<NOTE:1628298.1> - SuperCluster - ssctuner is not adjusting the scheduling class of all of lms , lgwr and vktm processes to FX-60

Attachments

This solution has no attachment