CPU over-commit in VMware cluster environment

I have read the note 1056052. The #5 guideline points out that SAP system should not use memory over-commit setting for virtual machines built on Windows with SQL server platform. My question is if this recommendation applys to CPU as well. Is it common to set CPU over commit for virtual machines that run dialog instance?
My environment is VMware vSphere 4.1, four ESX servers, ESX1 and ESX2 for production and ESX3 for DEV and ESX4 for QAS respectively. The initial placement is PRD CI+DB residents on ESX1 with all resources allocated, three PRD DIs each with 4 vCPU configured residents on ESX2 (total CPU capacity: 8, so CPU is over-committed here). Distributed Resource Scheduler (DRS) is enabled in the VMware cluster. I suppose DRS will initiate vMotion for one or two DIs from ESX2 to ESX3 or ESX4 with spare resources when PRD is loaded heavily.
Do you think the above system landscape and configuration is a practical one or reasonable? Thanks in adanvce for your help.

Matthias, thank you for your reply. You are right on the conclusion of CPU contention in case of high workload where CPU is over-committed. That's exactly where DRS can benefit SAP. I am now testing to see if CPU over-commitment will slow down the system in normal workload or the performance down is in an acceptable range. What's your insight of this?
At the same time, let's consider a different configuration. I remove the CPU over-commitment, for example 4 VMs each with 2 vCPU running on a 8-core ESX server. When the system is constantly stressed, DRS triggers vMotion of 1 VM off the specific ESX server. Comparing the response time before and after vMotion of that DI, I didn't see any improvement at all. Does this make sense? Can I conclude that DRS doesn't help in improving SAP performance whe CPU is not over-cmmitted on DI server?
Edited by: Fred Zhou on Nov 3, 2010 9:21 AM

Similar Messages

PI 7.1 in a cluster environment (multiple ip-adresses): P4 port

We want to install PI 7.1 on unix in a cluster environment.Therefore we installed also DEV+QA with virtual hostnames like the prodsystem, which will be later installed.
At all sapinst installation screens we have used only the virtual hostname <virtual-hostname-server interface>.We have also set the SAPINST_USE_HOSTNAME=<virtual-hostname-server interface>. Although the P4-port seems to have used the physical hostname: in step 57 of sapinst we got problems and in dev_icm were:
[Thr 05] *** ERROR => client with this banner already exists:
1:<physical-hostname>:35644 {000306f5} [p4_plg_mt.c 2495]
After we have set
icm/server_port_1 = PROT=P4,PORT=5$$04, HOST=<virtual-hostname-server interface>
icm/server_port_6 = PROT=P4,PORT=5$$04, HOST=<virtual-hostname-user interface>
icm/server_port_7 = PROT=P4,PORT=5$$04, HOST=<physical hostname>
icm/server_port_8 = PROT=P4,PORT=5$$04, HOST=127.0.0.1
the sapinst was successfull.
Now we're not sure how to set these P4-parameters in our future productive cluster environment.
Our productive system PX1 will live in a HA environment, so we don't want to use the physical hostnames in any profile.
Our environment will look like:
HOST-A (<physical-hostname-A>):
<virtual-hostname-server interface>
<virtual-hostname-user interface>
HOST-B (<physical-hostname-B>):
Normally our prodsystem will live on Host-A (physical-hostname-A). All parameters should
only take the virtual hostname <virtual-hostname-server-interface>. During switchover the
virtual hostnames (server and user interface) will be taken over to HOST-B, while the physical
hostnames of HOST-A and HOST-B will stay like there are.
How do the parameters have to be set here ?
Have also the physical hostnames of both cluster nodes set in the
instance profile, e.g:
icm/server_port_1 = PROT=P4,PORT=5$$04, HOST=<virtual-hostname-server interface>
icm/server_port_6 = PROT=P4,PORT=5$$04, HOST=<virtual-hostname-user interface>
icm/server_port_7 = PROT=P4,PORT=5$$04, HOST=<physical-hostname-A>
icm/server_port_8 = PROT=P4,PORT=5$$04, HOST=<physical-hostname-B>
icm/server_port_9 = PROT=P4,PORT=5$$04, HOST=<localhost>
Any recommendations ? In note 1158626 is some infomation regarding P4 ports with multiple network interfaces, but it's not 100% clear for us.
Best regards,
Uta

Hi Uta!
Obviously we are the only human beings in the SAP community having this problem. Nevertheless let's give it another try with a - hopefully - more simple problem description (and maybe it will be helpful to copy and paste this description into the open SAP CSN also).
So here comes the scenario:
We have one physical host:
Physical hostname: physhost
Physical IP address: 1.1.1.1
On this physical host there is running one OS: SUN Solaris 10/SPARC
On top of this we have two virtual hosts where we install 2 completely independent PI 7.1 instances with separate virtual hostnames and separate virtual IP addresses and separate DB2 9.1 databases. That is this is not an MCOD installation.
Virtual Host no. 1 is PI 7.1 Development System:
Virtual hostname: virthostdev
Virtual IP address: 2.2.2.2
Java Port numbers: 512xx
Virtual Host no. 2 is PI 7.1 QA System:
Virtual hostname: virthostqa
Virtual IP address: 3.3.3.3
Java Port numbers: 522xx
With this constellation we face serious problems with the P4 port. Currently for example the JSPM for virthostdev does not start, because JSPM cannot connect to the P4 port.
In SAP note 1158626 we have learned that by default always the physical hostname/IP address is used to address the P4 port and that we have to configure instance profile parameter icm/server_port_xx to avoid this.
So how do we have to configure the instance profile parameter icm/server_port_xx for both systems to resolve these P4 port conflicts?
Additionally: Is it important to use distinct server port slot numbers xx in both systems?
Additionally: Is it possible to configure this parameter with hostnames instead of using IP addresses?
So far we have tried several combinations, but with each combination at least one or even both systems have problems with that f.... P4 port.
Please help! Thanx a lot in advance!
Regards,
Volker

Mix guest os in a vmware cluster

Hi All,
I am trying to figure out if there is performance hit due to mix type of guest os in a vmware cluster e.g. Linux and windows VM's deployed together.
Is there any best practice document on it ?
Thanks

Yes that's correct, situation is related to large cluster of VDI environment. As Tom mentioned that different OS types may not be able to take benefit of Transparent Page Sharing. Additionally, Linux and windows have different way of using timer interrupts to keep os time up-to-date. Refer page 14 and 15 of below document - http://www.vmware.com/files/pdf/Timekeeping-In-VirtualMachines.pdf Not sure it would present a significant drain on resources --

Cluster environment / OBIEE 10.1.3.4.1

Hi All,
Can anyone give me more information on Clustering
1> Can we make schedulers work in a cluster environment ?
2> Max number of servers in a cluster configuration ?
Thanks
Goutham

The Cluster Server allows up to 16 Oracle BI Servers in a network domain to act as a single server.
Schedulers can also work in a clustered configuration however only one instance will always be active...
Active Scheduler. An active Scheduler is a clustered Oracle BI Scheduler instance which actively processes Scheduler jobs. The Cluster Controller determines the active instance at runtime and notifies the Scheduler cluster of this instance.
Inactive Scheduler. An inactive Scheduler is a clustered Oracle BI Scheduler instance which is not actively processing Scheduler jobs but is ready to take over in the event of an active Scheduler failure. An inactive scheduler is idle at all other times.

TimesTen database in Sun Cluster environment

Hi,
Currently we have our application together with the TimesTen database installed at the customer on two different nodes (running on Sun Solaris 10). The second node acts as a backup to provide failover functionality, although right now only manual failover is supported.
We are now looking into a hot-standby / high availability solution using Sun Cluster software. As understood from the documentation, applications can be 'plugged-in' to the Sun Cluster using Agents to monitor the application. Sun Cluster Agents should be already available for certain applications such as:
# MySQL
# Oracle 9i, 10g (HA and RAC)
# Oracle 9iAS Application Server
# PostgreSQL
(See http://www.sun.com/software/solaris/cluster/faq.jsp#q_19)
Our question is whether Sun Cluster Agents are already (freely) available for TimesTen? If so, where to find them. If not, should we write a specific Agent separately for TimesTen or handle database problems from the application.
Does someone have any experience using TimesTen in a Sun Cluster environment?
Thanks in advance!

Yes, we use 2-way replication, but we don't use cache connect. The replication is created like this on both servers:
create replication MYDB.REPSCHEME
element SERVER01_DS datastore
master MYDB on "SERVER01_REP"
transmit nondurable
subscriber MYDB on "SERVER02_REP"
element SERVER02_DS datastore
master MYDB on "SERVER02_REP"
transmit nondurable
subscriber MYDB on "SERVER01_REP"
store MYDB on "SERVER01_REP"
port 16004
failthreshold 500
store MYDB on "SERVER02_REP"
port 16004
failthreshold 500
The application runs on SERVER01 and is standby on SERVER02. If an invalid state is detected in the application, the application on SERVER01 is stopped and the application on SERVER02 is started.
In addition to this, we want to fail over if the database on the SERVER01 is in invalid state. What should we have monitored by the Clustering Agent to detect an invalid state in TT?

Oracle licensing in a Vmware Cluster

My question is related with Oracle licensing in a Vmware Cluster.
We have two VMware cluster in our datacenter: the first one (8 hosts – 16 sockets) is for not Oracle VM and the second (2 host – 3 sockets) only for Oracle VM. Both are connected to the same IBM V7000 with zoning and LUN masking that guarantees isolation.
In the Oracle cluster, DRS and HA are deactivated and VMotion logs are monitored for license compliance.
In this isolated environment, Oracle pretend to license every socket to any host connected to the V7000, regardless of the cluster that are connected the host.
Do you know how similar implementations with Oracle have been resolved (in terms of licensing)?
Thanks in advance!

Definitely a useful article. Also useful is this one:
http://houseofbrick.com/managing-oracle-licensing-in-a-shared-storage-environment/
And the various Oracle licensing-related articles written by Michael Webster:
http://longwhiteclouds.com/tag/oracle/
http://longwhiteclouds.com/tag/licensing/
To me personally it boils down to the very simple contractual key element:
By contract, you have to license every physical processor of every physical server where Oracle was/is running (whether that was as a VM or physical is irrelevant).
Conversely this means that by contract, you don't have to pay a penny for any physical server that never ever had Oracle software running on itself, period.
It's amazing how Oracle reps can keep ****ing with their customers like this.

Move SAP ECC standalone system to High Availability cluster environment.

Hi,
I have requirement to move SAP ECC stand alone system to cluster environment.
OS - MS server 2003
DB - Oracle 10
What steps need to follow to perform this activity?
What down time required?
Please Help.
Regards
Amit

Hi Amit
1. You have to perform SAP Homogeneous system copy method from soure to target
Kindly refer the system copy guide
http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/00f3087e-2620-2b10-d58a-c50b66c1578e?QuickLink=index&…
refer the SAP note FAQ 112266
2. Before that check the PAM you present version supported Windows 20008 / Oracle 11 G?
refer the SAP link http://service.sap.com/PAM
3. Over all down time may be 20 hrs time
BR
SS

SBO Mailer in a cluster environment

Hi,
I am currently facing an issue where when I install SBO Mailer on a cluster environment, the installation removes applications and license manager for some reason gets uninstalled.. When I want to uninstall the SBO Mailer, I do not see it under the Control Panel and neither do I see it in running in services. But if i look at the Service Manager the SBO Mailer Service is running.
The strange part is that after this installation, there is an add-on that stopped working.
Has anyone come across such a scenario?

Hi,
To uninstall you have to first run the Server tools and there you can uncheck the sbo mailer. There is not way you can uninstall SBO mailer service and you cannot find it under control panel its a service...not an applicaiton.
Run the server tools and then choose repair.
Regards,
Rakesh N

Deploying Java Web Application (WAR-File) into a cluster environment

Hi,
we have a web application which has to read from and write to the file system.
Since a short time we have a cluster environment (2 parallel servers) and since thisa time we have the problem, that files are worked double in the cluster. The application is working on both servers now and so we have this problem.
Does anybody know how we have to deploy the application correctly in a cluster environment or do we have to change anything in our source code of the application?
I didn't find any documentation about this.
At the moment we have deployed the application on one of the two servers only, but I think there must be a better way to solve this problem.
Thanks for your replies.
Regards
Thorsten

Hi,
I think first you need to wrap it into an EAR file, then you can deploy it.
As far as I know standalone deployment of WAR is deprecated as of 640.
similar threads:
How to deploy .war on NWDI
Deploying an existing WAR file into the Portal
Hopefully this tutorial also gives some idea:
http://help.sap.com/saphelp_nw70ehp1/helpdata/en/70/13353094af154a91cbe982d7dd0118/frameset.htm
Regards,
Ervin

Steps to upgrade kernel patch in AIX cluster environment

Hello All,
We are going to perform kernel upgrade in AIX cluster environment.
Please let me know the other locations to copy the new kernel files ,
default location
CI+DB server
APP1
Regards
Subbu

Hi Subbu
Refer the SAP link
Executing the saproot.sh Script - Java Support Package Manager (OBSOLETE) - SAP Library
1. Extract the downloaded files to a new location using SAPCAR -xvf <file_name> as sidadm.
2. copy the extracted files to sapmnt/<SID>/exe
3. Start the DB & Application.
Regards
Sriram

SAP R/3 4.7 SR1 Migration to MSCS cluster environment

Hi all,
We are planning to move our existing standalone server R/3 production server to cluster environment.
I have some queries regarding this migration
Our R/3 system is running on SAP R/3 4.7x1.10 SR1 on Windows 2003 and Oracle 9.2
1. Central Instance has to be installed in local disk or shared disk
2. Is it correct to R3load procedures to export data from old server and import during Database installation
3. Presently our server is running with kernel 640, but for the new installation kernel will be 620. So when can i upgrade kernel immediately after central instance installation or after CI and DB Instance installation.
Give me some inputs
Thanks & Regards
KT

Hi, the best place for your questions would have been the forum SAP on Windows.
However, here the answers.
1) CI must be on a shared disk
2) Right, if you don't change the database platform and its release you can also use backup/restore
3) Right. You can also perform the installation with the latest NW04 installation master

OSB Polling in Cluster Environiment

Hi ,
I created a OSB Poller which will read data from DB. When I test it in a single server it is working fine but when I move it to cluster environment with two managed servers it is polling Twice from the DB and the reports are generating twice . I have changed the JCA Property <Lock>lock-no-wait</Lock> even then it is polling twice. Any solution Please ?

Hi:
Maybe this link can be useful
http://javaoraclesoa.blogspot.com.au/2012/04/polling-with-dbadapter-in-clustered.html
Regards,
RK
Edited by: RK.. on Dec 16, 2012 7:54 PM
Edited by: RK.. on Dec 16, 2012 7:55 PM

How to setup the cluster environment for BPM using weblogic

want to setup the cluster environment for BPM using weblogic....
i have installed the oracle weblogic server 10gr3 and oracle BPM enterprise for weblogic 10gR3
i have used the Admin tools from the "oracle BPM enterprise for weblogic" to setup the configuration and create the weblogic domain servers.
i can launch the process administrator and import the project exp file to domain server.
but what should i do to setup cluster environment using weblogic?
what i want to do is :
setup one admin machine..
setup two product machine..
enable the cluster so the admin machine can monitor the status of the product machine..
thanks a lot ...

The install guide at http://download-llnw.oracle.com/docs/cd/E13154_01/bpm/docs65/config_guide/index.html gives a reasonable amount of info on how to do this.
Personally I have not used the OBPM option to configure WebLogic instead I've used the information in the above install guide to create the weblogic domain in advance of configuring OBPM.
Once you've setup WebLogic configure OBPM using the values I mention in the following thread: How to set the JMX Engine parameter in Process Administation?
Let me know any specific config questions and I'll do my best to answer them for you.
Thanks,
Mike

Changing the listener port number in a cluster environment

Hello,
I have an Oracle 10g database on a Windows cluster environment with Oracle Fail safe. I am trying to change the default listener port number - these are the steps I have done to change the port number:
1) Take the listener offline via Oracle Fail safe
2) stop the original listener from the command line
3) change the port number in the listener.ora file & save
4) start the original listener
5) bring the listener online in Fail safe
6) register the listener in the database with ALTER SYSTEM SET LOCAL_LISTENER....
After all this, when i check the status of the listener via lsnrctl, i see that the new port number is used, however in the Fail safe administrator, I still see the default port 1521. How do I go about changing the port number so that Fail safe also registers the change?

I did troubleshooting to verify the group, but this just changed the port number back to the default in the listener.ora & tnsnames.ora.
So I did all the steps again to change the port number from the default to another - via lsnrctl status, i see that the new port number is being used, I can also log in to the database via Toad using the new port number, in v$parameter i see that the local_listener is registered on the new port number....only under the Fail Safe manager, the port number (under listener parameter) has not changed....it still shows the default port number. Anyone know how to change this???

Installation of CRM 2007 in Windows with oracle and cluster environment

Dear Experts,
We are about to start the installation of CRM 2007 (ABAP+JAVA) with
Oracle 10g on Windows x64 in cluster environment. In the SAPINST dialog
box under High availability option, I could see installation options like ASCS
Instance, SCS Instance, First MSCS node, Database Instance, Additional
MSCS Node, Enqueue Replication Server, Central Instance and Dialog
Instance.
I have gone through the installation guide. I have below queries
regarding the same. Can you please clarify the same
1) Our requirement is we want to have an ACTIVE-ACTIVE cluster setup
with sap service running in Node A and Database running in Node B. Can
we have this setup in ABAP+JAVA setup
2) Also,in the SAPINST dialog box as said above except last two
(Central and Dialoge instance) as per standard installation guide are
to be installed in shared drives as per the requirement. But, central
and dialog are said to be installed locally on any one node in MSCS or
in separate server. As per my understanding Dialog instance will be
used for load balancing which we do not require now. Hence I feel this
is optional in our case. Where as Central Instance I am not able to
understand why this option is for? Is it mandatory to be installed or optional. If
so, when I installed it in one of the MSCS node the incase of failover
how does it effect the system. As per my understanding ASCS and SCS
comprise central instance.
Please clarify and thanks in advance.
Regards,
Sharath Babu M
+91-9003072891

I am following as per the standard installation guide.
Regards
Sharath

CPU over-commit in VMware cluster environment

Similar Messages

Maybe you are looking for