cancel
Showing results for 
Search instead for 
Did you mean: 

SF CFS for RAC without GCO option

shahfar
Level 5
Accredited Certified

Hi Everyone,

I have a cluster setup where I have a 2 node primary SF CFS for RAC cluster (for RHEL) and a single node as DR for the same. I am suing VVR to replicate the volumes accross to DR. I do not have the GCO option license. My question is, what should be the recommended VCS resources to be added to the configuration if I am not using GCO?

I have attached my main.cf file that shows my current config. Any suggstions would be appreciated. Thank you.

WBR

1 ACCEPTED SOLUTION

Accepted Solutions

mikebounds
Level 6
Partner Accredited

Just to clarify some permutations:

You can have SFCFS with failover Oracle - here Oracle only runs on one node at a time (controlled by VCS Oracle resources), so don't need an Oracle RAC license

SFCFS with Parallel Oracle RAC - here Oracle runs on both nodes in parallel, and this can be done in Solaris, but not sure about Linux (you didn't used to be able to do this with Linux) and I THINK even though you are using SFCFS (as oppose to Oracle Filesystem) you can choose whether VCS or CRS controls Oracle.

Mike

View solution in original post

8 REPLIES 8

Gaurav_S
Moderator
Moderator
   VIP    Certified

Hello,

If you are not using GCO .. that leaves VVR control of site in your hands (i.e manual control) ..

You can configure normal cluster resources on Prod site (no need of any RVG resources) & control VVR according to situation..

Also, you can read through RDC section from VCS admin guide, not sure though if that will suit your exact requirement..

https://sort.symantec.com/public/documents/sf/5.1/linux/html/vcs_admin/ch_vcs_replicated_data_clusters.html#221218

you can also find some sample main.cf files in the admin guide ..

 

G

mikebounds
Level 6
Partner Accredited

You should leave in the RVG resources you have already (RVGShared, RVGLogowner), but you should not add RVGSharedPri as this controls which node is the primary and you don't want the 2 clusters doing this if their is no coordination between them (i.e no GCO).

So you should leave config as it is, which means if you want to invoke DR you just need to run "vradmin takeover/migrate" before onlining the groups at the DR site.

Mike

shahfar
Level 5
Accredited Certified

Hi Guys,

Thanks for the input/suggestions.

Gaurav: Am I correct in assuming that for SF CFS for RAC on Linux, oracle resources are not controlled by VCS?

Mike: I do not have the SharedPri resource configured but I have configured CFS mount resources on both sites. The mount resrouce will be brought offline and then the replication roles will be changed manually. I have not configured the Oracle resources in VCS in the cluster. Have you had a chance to see my config files or load them in the simulator?

 

Thanks!

Gaurav_S
Moderator
Moderator
   VIP    Certified

yes you are correct , SFCFS for linux doesn't have full SFRAC infrastructure beneath .. in that case Oracle resources won't be controlled under VCS unless you explicitly create service groups with Oracle resource configured in it ..

 

G

mikebounds
Level 6
Partner Accredited

I had a quick look at config to see you had RVGShared, RVGLogowner resources but no RVGSharedPri which is correct.  Didin't notice you didn't have Oracle resources so you should add these, unless you don't want them controlled by VCS (I can't remember all the permuations of VCS and Oracle CRS control)

Process to failover would be offline oracle and CFS mount resources at Prod, sort out replication roles and then online CFS mount and Oracle resources at DR.

Note VVR will not allow filesystems to be mounted at secondary site, so it is safe (unlike an RVGPri resource) to have mount (and Oracle resources) configured.  However, if mounts resources auto start, they will fail at one site and if they don't autostart, you will need administrive intervention on a reboot.  You can sort this out with some triggers (i.e set groups to autostart, but have a preonline which prevents SG onlining if RVG is not primary)

Mike

mikebounds
Level 6
Partner Accredited

Just to clarify some permutations:

You can have SFCFS with failover Oracle - here Oracle only runs on one node at a time (controlled by VCS Oracle resources), so don't need an Oracle RAC license

SFCFS with Parallel Oracle RAC - here Oracle runs on both nodes in parallel, and this can be done in Solaris, but not sure about Linux (you didn't used to be able to do this with Linux) and I THINK even though you are using SFCFS (as oppose to Oracle Filesystem) you can choose whether VCS or CRS controls Oracle.

Mike

shahfar
Level 5
Accredited Certified

Hi Mike,

Thanks for your input and explanations. As far as I remember, Oracle on Linux does not support VCS handling Oracle resources for Oracle 10g. I am not sure, but I think this was the reason why Symantec came up with SF CFS for Linux instead of SF RAC on Linux. Although, I think I could pass on the Oracle service control to VCS as well but the DBA at this site does not want any unsupported configuration.

Additionally, on the DR site we have a single Oracle instance configured on one node as failover.

 

Thanks!

shahfar
Level 5
Accredited Certified

Hi Guys,

Although this thread is a bit old now, I still have one more question on this setup. The virtual IP that is used for replication between the primary and DR, how will I be able to manage that?

Lets say, on the primary cluster node 1 is the primary node for replication and the virtual IP is configured on this node. What happens when this node fails (In this case, node 2 (of the primary cluster) is still active)?

Will I have to manually bring up the same virtual IP used for replication on node 2?

Thanks!