Starting the Cluster

gpstart -a          # bring the entire cluster online
gpstop -r           # restart after configuration changes that require reboot
gpstop -u           # reload config files without stopping services

Master-only Maintenance Mode

gpstart -m          # start master in utility mode
PGOPTIONS='-c gp_session_role=utility' psql postgres   # connect for catalog work
gpstop -mr          # return to normal production mode

Shutting Down Services

gpstop              # graceful shutdown
gpstop -M fast      # immediate shutdown

Fialover to the Standby Master

gpactivatestandby -d $MASTER_DATA_DIRECTORY
gpstate -f          # verify standby promotion
mv /data/master/gpseg-1 /data/master/gpseg-1.bak      # archive old master dir

Daily Health Checks

gpstate                                     # overall cluster health
select * from gp_toolkit.gp_disk_free;      # free space from SQL
gpssh -f host.list -e "df -h | grep data"   # free space from OS

psql -c "select * from pg_stat_activity;"   # active connections
psql -c "select pg_size_pretty(pg_database_size('test'));"   # db size
psql -c "select pg_size_pretty(pg_relation_size('schema.test'));"  -- table size

select relname from pg_class where relname like 'ods%';
select gp_segment_id, count(*) from test group by 1;  -- data skew
select * from gp_segment_configuration order by 1;      -- segment config
vacuum analyze tablename;                             -- reclaim & stats

Monitoring Queries and Locks

select * from pg_stat_activity where waiting = 't' order by current_query;
select * from pg_stat_activity where state <> 'idle' and query_start < now() - interval '1 hour' order by query_start;

Segment Recovery Workflow

# 1. Check failed segments
gpstate -e

# 2. Standard recovery
nohup gprecoverseg -a &

# 3. Rebalance primaries if roles flipped
nohup gprecoverseg -ra &

# 4. Full rebuild when delta is too large
nohup gprecoverseg -Fa &

Transaction Control

BEGIN;                       -- or START TRANSACTION
-- DML statements
COMMIT;                      -- or END
-- or
ROLLBACK;

SAVEPOINT sp1;
-- more work
ROLLBACK TO SAVEPOINT sp1;
RELEASE SAVEPOINT sp1;

Free-Space Map Tuning

Expired rows are tracked in the free-space map. Tune with:

max_fsm_pages
max_fsm_relations

When the map overflwos, use CREATE TABLE AS to rebuild the table instead of VACUUM FULL.

Statistics and Vacuum

VACUUM ANALYZE tb_cp_02;

Log Management

Server logs roll daily and live in each instance’s pg_log directory:

gpdb-YYYY-MM-DD_TIME.csv

Verbosity levels:

show log_min_messages;     -- server side
show client_min_messages;  -- client side

Search logs:

gplogfilter -n 3                                    # last 3 lines on master
gpssh -f seg_host_file -e 'gplogfilter -n 3 /data/primary/*/pg_log/gpdb*.csv'

Utility logs are written to ~/gpAdminLogs/script_name_YYYYMMDD.log.

System Catalogs

Standard PostgreSQL catalogs live in pg_catalog. Greenplum additions include:

• gp_configuration
• gp_distribution_policy
• gp_distributed_log
• gp_id
• gp_version_at_initdb
• gp_master_mirroring
• gp_pgdatabase

\dtS   # list system tables
\dvS   # list system views