Comments on: RAC workshop http://laurentschneider.com/wordpress/2007/01/rac-workshop.html Oracle Certified Master Mon, 01 Dec 2008 18:34:47 +0000 http://wordpress.org/?v=2.6.5 By: Neeraj Bhatia http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-5067 Neeraj Bhatia Wed, 24 Oct 2007 03:14:27 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-5067 Its quite impressive text on RAC. very simple language and wonderfull explanation. Just bookmark the blog .... Thanks !! Its quite impressive text on RAC. very simple language and wonderfull explanation. Just bookmark the blog ….

Thanks !!

]]> By: Laurent Schneider http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4429 Laurent Schneider Mon, 30 Jul 2007 17:47:15 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4429 yes, you can use OEM or the views. Also the alert log. But I cannot imagine the instance recovery taking so long. You could start with the GV$MTTR_TARGET_ADVICE Can you connect and create a dummy table with sqlplus during those 25 minutes? yes, you can use OEM or the views. Also the alert log. But I cannot imagine the instance recovery taking so long.

You could start with the GV$MTTR_TARGET_ADVICE

Can you connect and create a dummy table with sqlplus during those 25 minutes?

]]> By: Irfan Khan http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4428 Irfan Khan Mon, 30 Jul 2007 17:38:10 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4428 Everything in weblogic log looks good ( from the log ). Its just the response time for the DB that is very very slow. And due to slow reponse times the users are being lost. I have a case open with weblogic, and we have multipools setup as they have recommended. I was going in the direction where there is a lot of RAC wait ( or some crash recovery on rac ) that is taking time. Is there a way to find out all the details of the recovery process and how long it takes Everything in weblogic log looks good ( from the log ). Its just the response time for the DB that is very very slow. And due to slow reponse times the users are being lost.

I have a case open with weblogic, and we have multipools setup as they have recommended.

I was going in the direction where there is a lot of RAC wait ( or some crash recovery on rac ) that is taking time.

Is there a way to find out all the details of the recovery process and how long it takes

]]> By: Laurent Schneider http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4427 Laurent Schneider Mon, 30 Jul 2007 17:26:02 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4427 you mean the weblogic multipool took the time? in this case you should check with weblogic support, because I cannot help you much there. But if you use taf, it should not take too long, and if it does, you can use PRECONNECT sessions. you mean the weblogic multipool took the time? in this case you should check with weblogic support, because I cannot help you much there.

But if you use taf, it should not take too long, and if it does, you can use PRECONNECT sessions.

]]> By: Irfan Khan http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4426 Irfan Khan Mon, 30 Jul 2007 17:15:36 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4426 I am sorry I did not explain the test clearly. We are crashing the node from the 2 node rac cluster. We want it to failover to the 2nd rac node (not the standby). I mentioned the standby coz I wanted to now if it will cause any delays. transfer is archive logs to the standby node (datagaurd) I am a Sys admin ( do a little bit of DBA ). We are testing multipools as oppose to TAF ( currently in production, taf going out of support by weblogic ) Hence we are testing the database server crash. We also tried a shutdown immediate onthe node A and node B took over immediately. NO loss of any users. But when we crash the database server it takes about 20-25 minutes to recovr, plus we loos a LOT of users I am sorry I did not explain the test clearly. We are crashing the node from the 2 node rac cluster. We want it to failover to the 2nd rac node (not the standby). I mentioned the standby coz I wanted to now if it will cause any delays.

transfer is archive logs to the standby node (datagaurd)

I am a Sys admin ( do a little bit of DBA ). We are testing multipools as oppose to TAF ( currently in production, taf going out of support by weblogic )

Hence we are testing the database server crash. We also tried a shutdown immediate onthe node A and node B took over immediately. NO loss of any users.

But when we crash the database server it takes about 20-25 minutes to recovr, plus we loos a LOT of users

]]> By: Laurent Schneider http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4425 Laurent Schneider Mon, 30 Jul 2007 16:42:20 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4425 what transfer do you use? if you use archivelog transfer, it will take a while... if you use log transfer (or online redo log hardware mirroring) it should never take that long, unless you use the DELAY parameter... I guess the instance recovery take only a few seconds, does not in? if not, you could set FAST_START_MTTR_TARGET, but I cannot believe this could be the root cause. Is the crash recovery slower because of RAC? or is it the same even if you have only 1 instance running what transfer do you use? if you use archivelog transfer, it will take a while…

if you use log transfer (or online redo log hardware mirroring) it should never take that long, unless you use the DELAY parameter…

I guess the instance recovery take only a few seconds, does not in? if not, you could set FAST_START_MTTR_TARGET, but I cannot believe this could be the root cause.

Is the crash recovery slower because of RAC? or is it the same even if you have only 1 instance running

]]> By: Irfan Khan http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4424 Irfan Khan Mon, 30 Jul 2007 14:00:55 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4424 We have a 2 node rac cluster and a standby ( datagaurd , single node rac ). We have 6 application servers connecting to the the 2 node rac cluster.( using weblogic multipools ). In our tests when we crash the databse server ( halt -q ) , it takes about 25 minutes to recover. I wanted to know how long is the oracle crash recovery process and what is so time consuming during this process? Also is datagaurd cozing additional delays thank you We have a 2 node rac cluster and a standby ( datagaurd , single node rac ). We have 6 application servers connecting to the the 2 node rac cluster.( using weblogic multipools ). In our tests when we crash the databse server ( halt -q ) , it takes about 25 minutes to recover.

I wanted to know how long is the oracle crash recovery process and what is so time consuming during this process? Also is datagaurd cozing additional delays

thank you

]]> By: Laurent Schneider http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4268 Laurent Schneider Fri, 13 Jul 2007 17:48:39 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4268 Jokach, Thanks for your question, you are very right to ask, it is not my habit to put such comments without justification. I may be completly wrong with some application that will scale well on your six nodes clusters. Also important is than you can offer different services :!: you have a 6 nodes database, but application A has 4 prefered nodes, application B has 4 prefered nodes and application C has 4 prefered nodes. So you keep a very high availability but you do not "scale" one application over more than 4 nodes, but you scale your whole cluster and add nodes as you offer more services/application. This could be an approach too. The concept about 5 nodes is less than 4 I have read a few times but could not find a reference right now, sorry. The fact that I have read it does not mean it is true :( But the overhead due to the cache fusion and the high number of connection will probably make your network a bottleneck if you add nodes and want to unlimitely scale your application. More tests should be needed. And it may be os/plateform/hardware specific Do not hesitate to ask more! ;-) Regards Laurent Jokach,
Thanks for your question, you are very right to ask, it is not my habit to put such comments without justification. I may be completly wrong with some application that will scale well on your six nodes clusters.

Also important is than you can offer different services :!: you have a 6 nodes database, but application A has 4 prefered nodes, application B has 4 prefered nodes and application C has 4 prefered nodes. So you keep a very high availability but you do not “scale” one application over more than 4 nodes, but you scale your whole cluster and add nodes as you offer more services/application.
This could be an approach too.

The concept about 5 nodes is less than 4 I have read a few times but could not find a reference right now, sorry. The fact that I have read it does not mean it is true :( But the overhead due to the cache fusion and the high number of connection will probably make your network a bottleneck if you add nodes and want to unlimitely scale your application. More tests should be needed. And it may be os/plateform/hardware specific

Do not hesitate to ask more!

;-)

Regards
Laurent

]]> By: jokach http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4266 jokach Fri, 13 Jul 2007 15:57:13 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-4266 "But if you have Linux, RAC offers additional scalability. However, it is not guarantee that your cluster will scale over 4 nodes." Is this written somewhere out there? I've heard this mentioned, and see it in your article, but can't seem to find any qualitative information supporting it? “But if you have Linux, RAC offers additional scalability. However, it is not guarantee that your cluster will scale over 4 nodes.”

Is this written somewhere out there? I’ve heard this mentioned, and see it in your article, but can’t seem to find any qualitative information supporting it?

]]> By: laurentschneider http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-566 laurentschneider Thu, 18 Jan 2007 10:18:47 +0000 http://laurentschneider.com/wordpress/2007/01/rac-workshop.html#comment-566 it is not very different from single instance. When you issue your STARTUP after your crash, the database is not open until crash recovery is finished. I hope I can test it more accurately one day it is not very different from single instance. When you issue your STARTUP after your crash, the database is not open until crash recovery is finished. I hope I can test it more accurately one day

]]>