cancel
Showing results for 
Search instead for 
Did you mean: 

How to cluster after BO installation

Former Member
0 Kudos

Hello here,

I have 4 independent instances of BO with there own local filestore and local CMS DB (Machine1, Machine2, Machine3, Machine 4). I now want to cluster these 4 machines as NODE of a cluster. I have done clustering at the time of install.....how does one cluster CMS , after they are installed ?

CCM--> Update Data source of machine2, machine3, machine4 ( to CMS DB of machine1 ) and then change the cluster name ?

Please advise.

Thanks,

Sam

Accepted Solutions (1)

Accepted Solutions (1)

denis_konovalov
Active Contributor
0 Kudos

This topic is covered in multiple threads, in Admin guide and in several KB articles.

Former Member
0 Kudos

Hello Denis,

Thanks for the response. I did check admin guide and could find clustering at the time of new install (do not select SQL AW DB and then provide name of existing CMS for 3.1 ) and could not find after the fact the installation is done.....could you please point or provide the right steps to it .

Thanks,

Sam

Former Member
0 Kudos

Hello Denis,

Thanks for the response. The link below really helped me out with standalone install.

http://wiki.sdn.sap.com/wiki/display/BOBJ/Appendix+1+-+Deployment+-+how+to+add+another+CMS+-+Multipl...

Quick question on that :

1) So all I have to do is create a new SIA ( say for machine2 ..... NODE2SIA) and point this new "NODE2SIA" to CMS db setup by Machine1 ...correct ?

2) Also, I should have it "Create default servers on the new node"  ...as I want full stack of servers on NODE2.....correct ?

3) Can I give SAME SIA name as that of SIA name with standalone install or do they have to be different? In this case still I will not Check "Recreate SIA on localhost" as CMS DB will be different ?

4) Per the document , Go to the registry and delete your first node (TmpSIA) :

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\BOE120SIATmpSIA]

Is this still true for 3.1 SP5 ...I could see the SIA entry in the registry .. but wanted to verify that as Doc. mentions it for SP2. Would just deleting service using "SC Delete servicename" on windows suffice it ...or we still have to muck around with registry ?

Thanks again Denis,

Sam

denis_konovalov
Active Contributor
0 Kudos

sam shaw wrote:

1) So all I have to do is create a new SIA ( say for machine2 ..... NODE2SIA) and point this new "NODE2SIA" to CMS db setup by Machine1 ...correct ?

Pretty much.

sam shaw wrote:

2) Also, I should have it "Create default servers on the new node"  ...as I want full stack of servers on NODE2.....correct ?

Yes

3) Can I give SAME SIA name as that of SIA name with standalone install or do they have to be different? In this case still I will not Check "Recreate SIA on localhost" as CMS DB will be different ?

I usualy give new name and keep original SIA alone. Just disabled.

4) Per the document , Go to the registry and delete your first node (TmpSIA) :

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\BOE120SIATmpSIA]

I do not do this step, as I keep all my original SIA. For troubleshooting purposes. Simply disabling it from auto start is enough.

This applies to all post XIR2 deployments.

Former Member
0 Kudos

Thanks again Denis.

Your solution to not deleting old SIA bring one other point.... in my case here is the scenario..

1) I have BO R2 clusterd and I am upgrading from R2 to 3.1. I will have also clustered BO 3.1 as mentioned above. Now the problem is , when I perform upgrade from BOXI R2 to BOXI 3.1 SP5, I will loose all of my cluster information in BOXI 3.1 CMS DB. BUT all the 4 nodes of BOXI 3.1 SIA will point to the same CMS DB ( which was used to upgrade from BOXI R2 ).

In this case, all I have to do on NODE2, NODE3, and NODE4 ....is to keep same old SIA but just recreate them with all default servers Or will I have create NEW SIA's again with different name for NODE2, NODE3 and NODE4??

I understand it would have ben much better idea to upgrade first and then create cluster in BOXI 3.1 but they way the environments are setup here...I am forced to cluster first, then upgrade and then fix the broken cluster .

Please advise.

Thanks Denis,

Sam

denis_konovalov
Active Contributor
0 Kudos

What upgrade process are you using ?

I sure hope not the CMSDBCOPY process !

graham_sturmy
Active Participant
0 Kudos

I couldn't tell from your description.  Do you intend to consolidate the content of the 4 seperate XI 3.1 nodes into a single 4 node cluster.

In this case you would need to use the import wizard first to migrate the content of the 3 other nodes to the node which is running the DB you wish keep for the cluster.  Only when you have done that should you perform the clustering.

Regards,
Graham

Former Member
0 Kudos

Hello Denis,

Yes I am using CCM --> Configuration --> Copy data from another data source. I understand that is not the most optimized way to do upgrade ( but cant use IW...as my QA nd prod are not conected and cant talk to each other). Also , content is too large to use . BIAR files.

Please advise.

Thanks,

Sam

denis_konovalov
Active Contributor
0 Kudos

Incremental upgrade then is an option...

otherwise you practicaly can skip the clustering part as CMSDBCOPY will destroy your 3.1 cluster and you'll have to re-do it from scratch...

Former Member
0 Kudos

Thanks again Denis.....thats exactly I am forced to do ...Cluster, Upgrade, fix cluster ( re-build it).

In order to do that, would you recommend recreating same SIA on Node2, 3,4 or you would suggest creating brand new SIA on Node2,3 and 4.

OR DELETE the SIA NODE 2,3 and 4 first ( by Deleting the SIA) and then upgrade from XIr2 to XIr3 and then recrate the SIA with OLD SIA names or new New SIA names on NODE 2,3,4..... in your experience you think this is better/cleaner approach ?

Thanks,

Sam

denis_konovalov
Active Contributor
0 Kudos

In that scenario, re-creating the nodes would be better, as it'll just add the nodes into the CMS DB and you won;t need to clear left over nodes on those machines.

Former Member
0 Kudos

Thanks Denis.

In that case would you DELETE the SIA NODE 2,3 and 4 first ( by Deleting the SIA) and then upgrade from XIr2 to XIr3 and then recrate the SIA with OLD SIA names OR NO need to DELETE SIA first on NODE 2,3 and 4.

denis_konovalov
Active Contributor
0 Kudos

I'd stop them all, except one, do the cmsdbcopy, then recreate those that are not started with same names they have already..
no need to delete.

Former Member
0 Kudos

Hey Denis,

Other quick question to that ... I tried to run RepoScan against upgraded NODE1 (using CMSDBCOPY) .....the scan is been running for about 3 hours and then after that I get scanning (300K objects....true number) ...and has been running for about an hour now....is this typical for reposcan?

Thanks,

Sam

denis_konovalov
Active Contributor
0 Kudos

it depends on your hardware, number of objects, CMSDB performance and usage of the system at the time.

Former Member
0 Kudos

Hello Denis,

Thanks for the reply. From your experience

1) Would you recommend using RepoScan ( with or without -repair) on the cut-over weekend if it takes this long to run and has known issues with repair ?

2) I found that Objects CUID's were changing from XIR2 to Xi3.1 using CMSDBCOPY.... should they change ?

Thanks again,

Sam

denis_konovalov
Active Contributor
0 Kudos

On 1 - none of this should be done on cut-over weekend.

On cut-over weekend both original and new environment should be operational already.

2 - SI_ID's should change, CUID's should not.

Former Member
0 Kudos

Thanks Denis...

1) Good to know ... but even before cut-over are you fan of using RepoScan ( with or without -repair)

2) Strange thing.....my CUID's have changed and hence Opendocs are failing as well ( I did change Opendoc URL for 3.1 from XIR2)......any clue on this ?

Thanks again,

Sam

denis_konovalov
Active Contributor
0 Kudos

reposcan is a tool that is usefull. that's it.

are u sure you're talking about CUID's ?

Former Member
0 Kudos

Thanks Denis.

BTW, did you experience issue with scheduling of Webi doc. in cluster? FOr e.g. if Node 1 CMS schedules webi doc to AJS of Node 2 or vice versa ... the schedule says in pending mode or the job fails ?

But if , NODE1 CMS schedules on NODE1 AJS or NODE2 CMS on NODE AJS.....webi jobs run just fine.....any thoughts ? I am forcing this via , server groups OR by shutting down CMS or AJS of one node .

Thanks,

Sam

michael_jennings
Participant
0 Kudos

Hi Denis,

    I'm looking a documentation about "How to clustering BOXI3.1 on Unix enviroment" Do You know where I can find that  ?

Best Regards,

MJ

Answers (0)