Does anyone have and example of SA template for Veritas Cluster Server that you can share ? Or any examples of scripts for Veritas Cluser Server exspeically with switching service groups to another server?
Don’t have any internally in the dev team I’m afraid.
Can you explain a bit more about what your trying to achieve here, are you patching an OS or applications in the cluster, or performing some other maintenance task ?
Veritas has come up in clustering discussions in the past so interested in the scenario you have.
Cheers,
Martin
EndPoint automation (002).pdf (177.5 KB)
Yes I have attached the use case, use original and version 3 shows more details with commands that is required to run. What is missing is the automation process around the commands.
Step A.Create Automatic Group with a client setting or property that UID each node so you can step through node 1, 2, 3
MAIRT , IF YOU HAVE SOME EXPERIENCE WITH HOW PROCESS THROUGH THE 3 NODES IN ORDER USING AUTOMATION , I WOULD APPRECIATE THE KNOWLEDGE AS I AM USING THE STEP A PROCESS SO THAT I CAN SPECIFY THE NODES IN ORDER AND USE THAT FOR ROUNDROBIN SWITCHING OF SERVICE GROUPS , (SWITCH TO NODE 2 , SWITCH TO NODE 3, ETC)
I AM BUILDING THIS TODAY AND TESTING ON THURSDAY EVENING IF COMPLETE ENOUGH. YOU ARE WELCOME TO JOIN AND IF YOU CAN HELP WITH TASKS TODAY THAT I CANNOT FIGURE OUT TO DO AS I AM NOT A EXPERT UNIX SCRIPT WRITER THAT WOULD BE GREAT. I NEED TO LOOP THROUGH THE OUTPUT FROM THE SERVICE GROUPS AND SWITCH EACH ONLINE GROUP TO ANOTHER NODE AND THEN AT END SWITCH THE SERVICE GROUP BACK. IF YOU HAVE SOME EXPERIENCE WITH THAT , I WOULD APPRECIATE IT.
Three-node clusters
Active-Passive
Step 1 Save current online service groups on each node.
Config file for 3 node cluster
parital file
cluster pdxcluu200d (
UserNames
= { admin = flkHigKrmKjfLhlFljHf,
dba_op = CkiGkqKjlGliGhlFhfGggE }
Administrators
= { admin }
Operators
= { dba_op }
)
system pdxcluu210d (
)
system pdxcluu211d (
)
system pdxcluu212d (
)
group Oracle_dcp140 (
SystemList
= { pdxcluu210d = 0, pdxcluu211d = 1, pdxcluu212d = 2 }
AutoStartList
= { pdxcluu210d }
FailOverPolicy
= RoundRobin
TriggersEnabled
@pdxcluu210d = { POSTONLINE, POSTOFFLINE }
TriggersEnabled
@pdxcluu211d = { POSTONLINE, POSTOFFLINE }
TriggersEnabled
@pdxcluu212d = { POSTONLINE, POSTOFFLINE }EndPoint automation-screenshot (003).pdf (182.3 KB)
)
Path: /etc/VRTSvcs/conf/config/main.cf
hastatus –sum |grep hostname|grep
ONLINE|grep –v Shared
Output to a file the results
/opt/VRTS/bin/hastatus –sum |grep hostname|grep ONLINE|grep –v Shared
root@pdxcluu210d# hastatus -sum |grep hostname|grep ONLINE|grep -v Shared (ONLINE SERVICE GROUPS THAT NEED TO BE SWITCHED ROUNDROBIN TO OTHER NODES) (SWITCHED BACK AT END OF PROCESS BEFORE MOVING TO NEW NODE)
B Oracle_dcp140 pdxcluu210d Y N ONLINE
B Oracle_dcs102 pdxcluu210d Y N ONLINE
B Oracle_dcs107 pdxcluu210d Y N ONLINE
B Oracle_dcs112 pdxcluu210d Y N ONLINE
B Oracle_dcs113 pdxcluu210d Y N ONLINE
BOracle_dds109 pdxcluu210d Y N ONLINE
B Oracle_din193 pdxcluu210d Y N ONLINE
NOTE: output the result into a file
Step 2 Failover online service groups from
to-be-patched-node to the other two nodes.
hagrp –switch to
hagrp –switch to
/opt/VRTS/bin/hagrp –switch Oracle_dcp140 pdxcluu211d
/opt/VRTS/bin/hagrp –switch Oracle_dcs102 pdxcluu212d
use Round Robin for target host
Step 3 Shutdown server to single-user mode
Step 4 Start EndPoint process
Step 5 Patch
Step 6 Restart node
Step 7 Failback service groups from other nodes.
hagrp –switch to
This task is for OS Solaris 10 / 11. But there are RHEL 6 as well.
Trying to get some of the guys to take a look at this, will respond with any thoughts/comments…
Thanks and I will keep you posted on progress
One highlight is in the Path: /etc/VRTSvcs/conf/config/main.cf has the three nodes ID but the system list is not reliable for node=0, node=1 etc as that is relative to the service group
)
system pdxcluu210d (
)
system pdxcluu211d (
)
system pdxcluu212d (
)
group Oracle_dcp140 (
SystemList
= { pdxcluu210d = 0, pdxcluu211d = 1, pdxcluu212d = 2 }
I’m afraid don’t have really have Veritas clustering experience in the team but we’ll do our best to help how we can.
One thing I notice is that you will be doing patching in single user node. Unfortunately the Bigfix agent will loose communication with the server in single user mode. There is a suggested solution for this at the following link but it doesn’t appear to be tested or supported by Bigfix/IBM
http://www-01.ibm.com/support/docview.wss?uid=swg21506015
If you are looking at a method for each node to be able to determine the full list of servers in the round-robin list, Server Automation has the concept of a Global paramaters fixlet which could possibly be used to set this list as a parameter for each step in the plan.
Regards,
John
Thanks and I will be happy to share all of my SA tasks with you and the team and including my customer of which now I have two that want this SA process would be happy to assist I am sure. But I would have to inquiry if you choose so. We are using the tasks in BigFix from the ;Solaris site for
a.The Single-User Mode Task - Solaris(ID #28 in the Patches for Solaris site).b.One or more Solaris Patch Fixlets available in the Patches for Solaris site.c.Depending on the requirements of the patch, choose either:vReboot Task - Solaris(ID #32 in the Patches for Solaris site)vReconfigure Reboot Task - Solaris(ID #30 in the Patches for Solaris site.
They are working fine except for the Single Mode Task which this one provides restart of the agent but in Solaris Zone the agent attempts to restart twice and does not complete. All the cluster zones are experiencing this problem I will look into the Global Parameters. I will keep this post updated as I go and will reach out.
Thanks
You could check the way the agents are installed, configured, make sure thats as it should be…
https://www-01.ibm.com/support/knowledgecenter/SS63NW_9.2.0/com.ibm.tivoli.tem.doc_9.2/Platform/Adm/c_solaris_installation_instructi.html
I suspect the global parameter suggestion won’t work as the implementation we have today was developed for WAS, and I doubt it would be generic enough to handle custom content developed in the same style.
In other clusters we’ve usually had a vendor API to query info about other nodes, so without knowing the Veritas interface, is there some other API that might return a more consistent view of the nodes in the cluster, rather than the “relative” view you’re currently getting ?
By all means keep us informed how you get on, and whether other issues arise where we might be able to offer better help.
Thanks Mairt,
I had sent the same URL to the customer to verify installation process and they used this process. Thanks for the update and I will ask about the API. I will keep you posted.