BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
sasprofile
Quartz | Level 8

Hello friends,

 

Please help me

 

I need your assistance in Adding Additional SAS Application Server in our new SAS 9.4 environment.

I got a requirement to add a additional application server into a new SAS 9.4 environment.

This is the first time am doing this configuration It would really a big help to me if you can guide me.

and I did not find a specific doc in sas support which can help me with this configuration, I found a doc but it does not
provide enough information for this Application server configuration

 

our environment has Total six servers MetadataServer, Grid Master Server, Grid Node 1 Server, Grid Node 2 Server, Grid Node 3 Server, Midtier Server

 

we already have a default application server in SMC and since object spawner is already there so am using same object spawner

for the new Application server.

 

below are initial steps am performing to add the Application server.

 

Please correct me if am wrong with any of the below initial stepse, I would really appreciate if anyone can correct
me if am wrong.

 

1) First I will be Taking the Metadata backup from the SAS Management console

 

2) Second I will be stopping all the below Servers in the below order

 

Midtier Server

Grid Node 3 Server

Grid Node 2 Server

Grid Node 1 Server

Grid Master Server

Metadata Server

 

3) Next I will be taking the Config,SASHome, Install files backups from Metadata Server, Grid Master Server and Midtier Node

 

4) Next I will be restarting  the servers in the below order

 

MetadataServer

Grid Master Server

Grid Node 1 Server

Grid Node 2 Server

Grid Node 3 Server

Midtier Server


5) Then Next I will be running ./setup.sh file from the SAS Grid Master Server to launch Deployment wizard

When the Deployment wizard launches select the options as described below


a) First option Select Install Software


b) Second Select a previously created SASHome


c) Next select first radio button Perform a Planned Deployment
   under this uncheck Install SAS Software checkbox and select only Configure SAS Software


d) Next Select first radio button Specify the full path to a customized deployment plan


e) Next step under Deployment Step I see below options, from this If am not wrong I should select 2nd option
SAS Grid Control and LSF Master Server

 

SAS MetadataServer
SAS Grid Control and LSF Master Server
SAS Grid Manager Node Servers
SAS Middle Tier Server
SAS Middle Tier Server Node
SAS Clients
SAS Grid Manager Client

 

 

 

Thank you in Adavnace for all your help

 

 

1 ACCEPTED SOLUTION

Accepted Solutions
jklaverstijn
Rhodochrosite | Level 12

I will try and answer your questions to the best of my knowledge. This oculd become a long post. There is so much detail involved in a solid grid implementation that you may consider to call in the troops (from SAS or otherwise) to iron out the last wrinkles. There is a good chance my answers will give rise to even more questions. Please understand that this platform may not be the place for an in-depth grid implementation workshop. But I will give it a shot.


Is the SAS servers stop and start  order is correct  which I mentioned in my previous email? 

There is a chapter in the SAS docs that recommends a certain order. Yours looks fine. Some servers can be started at the same time (the grid nodes) but that's a minor detail. Actually the grid nodes do not run more than a tiny LSF service and the environment manager agent; no spawners live here. The evmgr agent can be stopped/started at will.

 


Do I need to take backup of all files including SASHome, platform(gmsinstall, Sharedgrid, lsf,pminstall...etc) along with config files of all servers(Metadataserver,Grid master & Midtier) (or) just grid master is enough since am running setup file on SAS grid master.

Taking a backup is (almost) always a good idea. In your specific case, adding a server context, You could limit yourself to a metadata backup taken from SAS Management Console (smc) just before lanching sdw. It's just so much easier to roll back in case of mistakes.



And when it asks for option in the windows clicking Next during my configuration do I need to select existing config lev1 which is in our current grid master location
 

Never just click next. Always check what's there. And yes, you use the same config/Lev{n} as before. I just commented on this casually as I think most screens prompt you for info that you have seen before and are not special in the context of your case. The screens are pre-filled from previous sdw runs and will usually make sense, just don't count on it.




and also in my requirement am supposed to setup the saswork and sasutil files for the new Appserver,where do i need to setup these,do i need to setup in SMC or in Server.
 

Usually (not always! You can do it at SASHome which many frown upon) work/util locations are defined in the sasv9_usermods.cfg at server context level (eg. SASConfig/Lev1/SASApp). You will have to redo that for the new server. Up to you if you need a new location or reuse the same. It largely depends on why you need a new server context if you want to separate these.


You mentioned in your previous email  saying that "We work according to the best practice of sharing the SASConfig between all grid nodes, in which case this would the only time you need to run the SDW". (so what exactly you mean when you are saying I need to run SDW)....For your information our Grid Master and Nodes are in NFS for Install and config directories
 

We also use NFS for that. And we mount the same SASConfig location on each and every grid node including the master. This works very well but the devil is in the details. Eg. the env mgr agent writes a file called .pid at  startup so each node overwrites that file fro the previous one. The environment manager will loose track of the status of the agents. That can be helped by splitting the config directory in one for each node. SAS Support should be able to provide a a script to help. Another consequence is that you can't run SDW to configure every grid node separately. So there is no way to add the workspace servers to the metadata but manually. Unless you use grid launched wss exclusively there is no escape.  By the way NFS proves very problematic for us. We are in the process of implementing GPFS for that.

 



Do I need to select OLAP and connect server also as part of the this new app server context configuration (or) I can use the previous OLAP and connect server components for the new SASApp server.
 
and coming to port numbers I think for OLAP and Connect server i should use unique port numbers not the same port numbers which we used in our default application server and port numbers for workspace servers and pooled wss and stored process sever I believe we can use the same previous port numbers or new port numbers if am not wrong.
 
but in case If i have to use new port numbers for all the server components of new SASApp Server,do I need to use the next sequence number of the existing port numbers  of server components in default application server or it's something different.
 

You can use the existing ones. Normally the sdw will propose new port numbers that increment in the last or one-but-last digit. Some planning upfront will never hurt but the minimal requirement is to keep them unique.


 and to Save a response file for later reference I believe below is the command right for the response file to
 
./setup.sh -deploy -record -responsefile /path to response file/respfile.txt
 

Yes that's how it's done. The filename should (on Unix/Linux at least) be an absolute path name.

 

I hope this helps you on your way. The grid architecture can be a bit daunting in the beginning but the rewards are many. Good luck.

 

Regards,

- Jan.

 

View solution in original post

7 REPLIES 7
jklaverstijn
Rhodochrosite | Level 12

Hi,

 

You're on the right track. Just push forward from here. Choose the grid control server. Say yes when the wizard warns for using a config directory that already contains files. In the next step you get to select the products. Start with deselecting all and then select the server components you wish to add: Workspace Server, Pooled WSS, etc. And indeed, a new object spawner is not needed. Select a series of Next as they will be obvious to you, connecting you to the proper metadata server etc. The next crucial choice will be to "Create a SAS application server context". Check this option and click Next. Then choose a new appserver context name (different from the existing one, like SASApp). From here onwards everything is the same as before. So let me know if there are questions.

 

We work according to the best practice of sharing the SASConfig between all grid nodes, in which case this would the only time you need to run the SDW.

 

Tip: Save a response file for later reference and input for yet another server context.

 

It's as easy at that. Hope you will have smooth sailing.

 

Regards,

- Jan.

sasprofile
Quartz | Level 8

Hi Jan,

 

Thanks alot for the prompt reply.

 

I have few doubts I want to clarify with you

 

Is the SAS servers stop and start  order is correct  which I mentioned in my previous email?

 

Do I need to take backup of all files including SASHome, platform(gmsinstall, Sharedgrid, lsf,pminstall...etc) along with config files of all servers(Metadataserver,Grid master & Midtier) (or) just grid master is enough since am running setup file on SAS grid master.
 
And when it asks for option in the windows clicking Next during my configuration do I need to select existing config lev1 which is in our current grid master location
 
 
and also in my requirement am supposed to setup the saswork and sasutil files for the new Appserver,where do i need to setup these,do i need to setup in SMC or in Server.
 
You mentioned in your previous email  saying that "We work according to the best practice of sharing the SASConfig between all grid nodes, in which case this would the only time you need to run the SDW". (so what exactly you mean when you are saying I need to run SDW)....For your information our Grid Master and Nodes are in NFS for Install and config directories
 
Do I need to select OLAP and connect server also as part of the this new app server context configuration (or) I can use the previous OLAP and connect server components for the new SASApp server.
 
and coming to port numbers I think for OLAP and Connect server i should use unique port numbers not the same port numbers which we used in our default application server and port numbers for workspace servers and pooled wss and stored process sever I believe we can use the same previous port numbers or new port numbers if am not wrong.
 
but in case If i have to use new port numbers for all the server components of new SASApp Server,do I need to use the next sequence number of the existing port numbers  of server components in default application server or it's something different.
 
and to Save a response file for later reference I believe below is the command right for the response file to
 
./setup.sh -deploy -record -responsefile /path to response file/respfile.txt
 
I Would appreciare if you can clarify my above doubts
 
Thank you

 

jklaverstijn
Rhodochrosite | Level 12

I will try and answer your questions to the best of my knowledge. This oculd become a long post. There is so much detail involved in a solid grid implementation that you may consider to call in the troops (from SAS or otherwise) to iron out the last wrinkles. There is a good chance my answers will give rise to even more questions. Please understand that this platform may not be the place for an in-depth grid implementation workshop. But I will give it a shot.


Is the SAS servers stop and start  order is correct  which I mentioned in my previous email? 

There is a chapter in the SAS docs that recommends a certain order. Yours looks fine. Some servers can be started at the same time (the grid nodes) but that's a minor detail. Actually the grid nodes do not run more than a tiny LSF service and the environment manager agent; no spawners live here. The evmgr agent can be stopped/started at will.

 


Do I need to take backup of all files including SASHome, platform(gmsinstall, Sharedgrid, lsf,pminstall...etc) along with config files of all servers(Metadataserver,Grid master & Midtier) (or) just grid master is enough since am running setup file on SAS grid master.

Taking a backup is (almost) always a good idea. In your specific case, adding a server context, You could limit yourself to a metadata backup taken from SAS Management Console (smc) just before lanching sdw. It's just so much easier to roll back in case of mistakes.



And when it asks for option in the windows clicking Next during my configuration do I need to select existing config lev1 which is in our current grid master location
 

Never just click next. Always check what's there. And yes, you use the same config/Lev{n} as before. I just commented on this casually as I think most screens prompt you for info that you have seen before and are not special in the context of your case. The screens are pre-filled from previous sdw runs and will usually make sense, just don't count on it.




and also in my requirement am supposed to setup the saswork and sasutil files for the new Appserver,where do i need to setup these,do i need to setup in SMC or in Server.
 

Usually (not always! You can do it at SASHome which many frown upon) work/util locations are defined in the sasv9_usermods.cfg at server context level (eg. SASConfig/Lev1/SASApp). You will have to redo that for the new server. Up to you if you need a new location or reuse the same. It largely depends on why you need a new server context if you want to separate these.


You mentioned in your previous email  saying that "We work according to the best practice of sharing the SASConfig between all grid nodes, in which case this would the only time you need to run the SDW". (so what exactly you mean when you are saying I need to run SDW)....For your information our Grid Master and Nodes are in NFS for Install and config directories
 

We also use NFS for that. And we mount the same SASConfig location on each and every grid node including the master. This works very well but the devil is in the details. Eg. the env mgr agent writes a file called .pid at  startup so each node overwrites that file fro the previous one. The environment manager will loose track of the status of the agents. That can be helped by splitting the config directory in one for each node. SAS Support should be able to provide a a script to help. Another consequence is that you can't run SDW to configure every grid node separately. So there is no way to add the workspace servers to the metadata but manually. Unless you use grid launched wss exclusively there is no escape.  By the way NFS proves very problematic for us. We are in the process of implementing GPFS for that.

 



Do I need to select OLAP and connect server also as part of the this new app server context configuration (or) I can use the previous OLAP and connect server components for the new SASApp server.
 
and coming to port numbers I think for OLAP and Connect server i should use unique port numbers not the same port numbers which we used in our default application server and port numbers for workspace servers and pooled wss and stored process sever I believe we can use the same previous port numbers or new port numbers if am not wrong.
 
but in case If i have to use new port numbers for all the server components of new SASApp Server,do I need to use the next sequence number of the existing port numbers  of server components in default application server or it's something different.
 

You can use the existing ones. Normally the sdw will propose new port numbers that increment in the last or one-but-last digit. Some planning upfront will never hurt but the minimal requirement is to keep them unique.


 and to Save a response file for later reference I believe below is the command right for the response file to
 
./setup.sh -deploy -record -responsefile /path to response file/respfile.txt
 

Yes that's how it's done. The filename should (on Unix/Linux at least) be an absolute path name.

 

I hope this helps you on your way. The grid architecture can be a bit daunting in the beginning but the rewards are many. Good luck.

 

Regards,

- Jan.

 

sasprofile
Quartz | Level 8

Jan-


You are saying to take just only a Metadat backup from SMC,and nothing else.
but normally when we do any new configuration we are supposed to take backup of the SASHome,Install and config files right,In case if the new configurationmess up any of the existing configuration then we can roll back.
so Is it not necessary to take backup of the SASHome, Install and Config files on all servers Metadata Server, Grid Master Server and Midtier Server.

 

and For saswork and sasutil,so I can setup the work and util files for the new SASAppserver context in SASConfig/Lev1/NewSASAppserver/sasv9_usermods.cfg

 

Thank you

 

Alex888
Fluorite | Level 6

Hi Jan ,

 

In the SAS document on start/stop order SAS Service Start/Stop Order , which you mentioned earlier, looks like SAS does not consider GRID factor, the document does not mention anything related with grid enviornment.

 

My question is:  in a GRID enviormnent, should the Grid Control Server be started right after starting the Metatdata Server, and before the other Compute Servers (i.e., regular Grid Nodes)? And when stopping the servers, should the Grid Control Server be stopped just before stopping the Metadata server but after stopping all other grid nodes ?

 

I appreciate your prompt !

 

sasprofile
Quartz | Level 8

Jan-

You are saying to take just only a Metadat backup from SMC,and nothing else.
but normally when we do any new configuration we are supposed to take backup of the SASHome,Install and config files right,In case if the new configurationmess up any of the existing configuration then we can roll back.
so Is it not necessary to take backup of the SASHome, Install and Config files on all servers Metadata Server, Grid Master Server and Midtier Server.

and For saswork and sasutil,so I can setup the work and util files for the new SASAppserver context in SASConfig/Lev1/NewSASAppserver/sasv9_usermods.cfg

Thank you
jklaverstijn
Rhodochrosite | Level 12

Yes you understood right. Adding a server context adds metadata and a subdirectory under Lev{n} in your compute tier. So no need to do dramatic things to middle tier. In the SDW you choose Configure but not Install so the SASHome is not affected at all. Files on the metadata server or niddle tier are not impacted either. So your reguylar backups will suffice; no need to do anything extra here.

 

As I explained before we have a shared config for grid master and grid nodes so here we have one backup to rule them all.

 

Having said that, if you prefer to adhwere to your current practices that by all means don't let me stop you. Better safe than sorry, right?

 

And yes again on your understanding of where to configure the locations of work and util.

 

Regards, Jan.

suga badge.PNGThe SAS Users Group for Administrators (SUGA) is open to all SAS administrators and architects who install, update, manage or maintain a SAS deployment. 

Join SUGA 

Get Started with SAS Information Catalog in SAS Viya

SAS technical trainer Erin Winters shows you how to explore assets, create new data discovery agents, schedule data discovery agents, and much more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 7 replies
  • 5020 views
  • 1 like
  • 3 in conversation