Table of Contents

DB2 pureScale

Installation & Configuration

Linux pre-requisites : 

yum -y install

Change redhat release : 

Red Hat Enterprise Linux Server release 6.7 (Santiago)

Install DB2 on-all nodes !

Prepare /var/ct/cfg/netmon.cf : 

!IBQPORTONLY !ALL
!REQD eth0 192.168.88.10
!REQD eth3 192.168.56.1

Update existing instance to pureScale : 

./db2iupdt -cf db2dev01 -cfnet db2dev01-hadr -m db2dev01 -mnet db2dev01-hadr -mid 0 -instance_shared_dev /dev/sda -tbdev /dev/sdb -u db2pure1 db2pure1

Create new instance :

GPFS

Create GPFS FS Cluster

Create GPFS Cluster filesystem domain : 

db2cluster -cfs -create -domain db2cluster_TEST.unixguest.eu -host db2purescaleA

Set the GPFS Tiebreaker to disk device : 

db2cluster -cfs -set -tiebreaker -disk /dev/sdb

Start the GPFS Cluster : 

db2cluster -cfs -start -all

Create Instance shared FS : 

db2cluster -cfs -create -filesystem db2inst1_shared -disk /dev/sdc -mount /db2inst1/shared

List shared filesystems : 

db2cluster -cfs -list -filesystem

Stop GPFS filesystems : 

db2cluster -cfs -stop -all

Add new GPFS filesystem : 

db2cluster -cfs -create -filesystem db2fs_data_01 -disk /dev/sdb -mount /db2_DATA_01

Remove shared GPFS FS & GPFS Domain : 

# List the FS
db2cluster -cfs -list -filesystem

# Change the tiebreaker to Majority
db2cluster -cfs -stop -all
db2cluster -cfs -set -tiebreaker -majority
db2cluster -cfs -start -all

# Delete all shared files
db2cluster -cfs -mount -filesystem db2fs1
rm -rf </mount/point>/.*

# Remove the Shared FS
db2cluster -cfs -delete -filesystem db2fs1

# Remove the RSCT Domain
db2cluster -cfs -stop -all
db2cluster -cfs -list -domain
db2cluster -cfs -delete -domain <DOMAIN>

# Clear the DB2 glogal registry
db2greg -delvarrec service=GPFS_CLUSTER,variable=NAME,installpath=-
db2greg -delvarrec service=DEFAULT_INSTPROF,variable=DEFAULT,installpath=-

Enable PR-3 Reservations :

GPFS

Check GPFS cluster configuration : 

mmlscluster

Troubleshooting

If primary GPFS cluster fails, designate new primary server : 

mmchcluster -p <NEWPRIMARY> -s <NEWSECONDARY>

Cluster Ping tool : 

[[http://www-01.ibm.com/support/docview.wss?uid=swg21967473]]

Enable member node as Quorum node : 

export CT_MANAGEMENT_SCOPE=2
chrsrc -s 'Name="<NODENAME>"' IBM.PeerNode IsQuorumNode=1

Enable failback mode for RG : 

1. Log in as root from any cluster node
2. export CT_MANAGEMENT_SCOPE=2
3. rgreq -o lock db2_db2sdin1_0-rg
4. rgreq -o lock db2_db2sdin1_1-rg
5. chrsrc -s "MemberOf = 'db2_db2sdin1_0-rg'" IBM.ManagedResource SelectFromPolicy=3
6. chrsrc -s "MemberOf = 'db2_db2sdin1_1-rg'" IBM.ManagedResource SelectFromPolicy=3
7. rgreq -o unlock db2_db2sdin1_0-rg
8. rgreq -o unlock db2_db2sdin1_1-rg

Attaching to DB2 processes / generating core dumps : 


-Lock all member resource groups as root

rgreq -o lock db2_inst0_0-rg
rgreq -o lock db2_inst0_1-rg
rgreq -o lock db2_inst0_2-rg
rgreq -o lock db2_inst0_3-rg

-Gather a core file ( on Linux attach and gcore from gdb, on AIX by gencore )

-Unlock all members?as follows as root

rgreq -o unlock db2_inst0_0-rg
rgreq -o unlock db2_inst0_1-rg
rgreq -o unlock db2_inst0_2-rg
rgreq -o unlock db2_inst0_3-rg

Enable GPFS failover on multipath : 

http://www-01.ibm.com/support/docview.wss?uid=swg22003235&myns=swgimgmt&mynp=OCSSEPGG&mync=E&cm_sp=swgimgmt-_-OCSSEPGG-_-E

Kerberos authentication

  1. Create KRB database

kdb5_util create -s