Greenplum 7 Resource Group v2

Greenplum 7 워크로드 매니져, Resource Group v2

1. Greenplum 7에서는 워크로드매니저를 3가지 형태로 제공

1) Greenplum Resource Queue

- Default 이며, Greenplum 3버전 이후에 제공하는 기본 워크로드 매니저

- ACTIVE_STATEMENTS, MEMORY_LIMIT, MAX_COST, CPU PRIORITY로 관리

2) Greenplum Resource Group cgroup v1

- Greenplum 5 이후 지원되며, 리눅스의 cgroup v1 기반으로 워크로드 관리

- CONCURRENCY, CPU_MAX_PERCENT, CPU_WEIGHT, CPUSET, MEMORY_LIMIT, MIN_COST로 관리

3) Greenplum Resource Group cgroup v2:

- Greenplum 7 이후 지원되며, 리눅스의 cgroup v2 기반으로 워크로드 관리

- CONCURRENCY, CPU_MAX_PERCENT, CPU_WEIGHT, CPUSET, MEMORY_LIMIT, MIN_COST + IO_LIMIT로 관리

- Disk IO 컨트롤 까지 지원 (Disk Read/Write의 bps/iops)

2. Greenplum 7의 Resource Group v2 설정 방법

1) 메뉴얼

- https://docs.vmware.com/en/VMware-Greenplum/7/greenplum-database/admin_guide-workload_mgmt_resgroups.html

2) OS cgroup v2 설정 (모든 노드, root 계정)

# all

=> grubby --update-kernel=ALL --args="systemd.unified_cgroup_hierarchy=1"

[ cdw.gpdbkr.com]

[scdw.gpdbkr.com]

[sdw1.gpdbkr.com]

[sdw2.gpdbkr.com]

[sdw3.gpdbkr.com]

[sdw4.gpdbkr.com]

=> sync;sync;

=> reboot now

# all

[root@cdw ~]# all

=> grubby --info=DEFAULT | grep args

[ cdw.gpdbkr.com] args="ro crashkernel=auto ... $tuned_params systemd.unified_cgroup_hierarchy=1"

[scdw.gpdbkr.com] args="ro crashkernel=auto ... $tuned_params systemd.unified_cgroup_hierarchy=1"

[sdw1.gpdbkr.com] args="ro crashkernel=auto ... $tuned_params systemd.unified_cgroup_hierarchy=1"

[sdw2.gpdbkr.com] args="ro crashkernel=auto ... $tuned_params systemd.unified_cgroup_hierarchy=1"

[sdw3.gpdbkr.com] args="ro crashkernel=auto ... $tuned_params systemd.unified_cgroup_hierarchy=1"

[sdw4.gpdbkr.com] args="ro crashkernel=auto ... $tuned_params systemd.unified_cgroup_hierarchy=1"

=> stat -fc %T /sys/fs/cgroup/

[ cdw.gpdbkr.com] cgroup2fs

[scdw.gpdbkr.com] cgroup2fs

[sdw1.gpdbkr.com] cgroup2fs

[sdw2.gpdbkr.com] cgroup2fs

[sdw3.gpdbkr.com] cgroup2fs

[sdw4.gpdbkr.com] cgroup2fs

=> mkdir -p /sys/fs/cgroup/gpdb.service

=> echo "+cpuset +io +cpu +memory" | tee -a /sys/fs/cgroup/cgroup.subtree_control

[scdw.gpdbkr.com] +cpuset +io +cpu +memory

[sdw4.gpdbkr.com] +cpuset +io +cpu +memory

[sdw2.gpdbkr.com] +cpuset +io +cpu +memory

[sdw3.gpdbkr.com] +cpuset +io +cpu +memory

[ cdw.gpdbkr.com] +cpuset +io +cpu +memory

[sdw1.gpdbkr.com] +cpuset +io +cpu +memory

=> chown -R gpadmin:gpadmin /sys/fs/cgroup/gpdb.service

=> chmod a+w /sys/fs/cgroup/cgroup.procs

=> cat /sys/fs/cgroup/cgroup.subtree_control

[scdw.gpdbkr.com] cpuset cpu io memory pids

[sdw4.gpdbkr.com] cpuset cpu io memory pids

[sdw2.gpdbkr.com] cpuset cpu io memory pids

[sdw3.gpdbkr.com] cpuset cpu io memory pids

[ cdw.gpdbkr.com] cpuset cpu io memory pids

[sdw1.gpdbkr.com] cpuset cpu io memory pids

=> exit

########### 메뉴얼에 나온 부분을 일부 수정

########### 시스템 리부팅하고 난뒤에 gpstart가 되지 않을 경우 아래와 같이 수정

########### 변경 사항

WorkingDirectory=/sys/fs/cgroup/gpdb.service => WorkingDirectory=/sys/fs/cgroup

chown -R gpadmin:gpadmin .; \ => chown -R gpadmin:gpadmin ./gpdb.service; \

chmod a+w ../cgroup.procs; \ => chmod a+w ./cgroup.procs; \

########### gpdb.service 소스

[root@cdw ~]# vi /etc/systemd/system/gpdb.service

[root@cdw ~]# cat /etc/systemd/system/gpdb.service

[Unit]

Description=Greenplum Cgroup v2 Configuration Service

[Service]

Type=simple

WorkingDirectory=/sys/fs/cgroup

Delegate=yes

Slice=-.slice

# set hierarchies only if cgroup v2 mounted

ExecCondition=bash -c '[ xcgroup2fs = x$(stat -fc "%%T" /sys/fs/cgroup) ] || exit 1'

ExecStartPre=bash -ec " \

chown -R gpadmin:gpadmin ./gpdb.service; \

chmod a+w ./cgroup.procs; \

mkdir -p helper.scope"

ExecStart=sleep infinity

ExecStartPost=bash -ec "echo $MAINPID > ./helper.scope/cgroup.procs; "

[Install]

WantedBy=basic.target

########### 모든 노드에 gpdb.service 복사

[root@cdw ~]#

[root@cdw ~]# scp /etc/systemd/system/gpdb.service smdw:/etc/systemd/system/gpdb.service

[root@cdw ~]# scp /etc/systemd/system/gpdb.service sdw1:/etc/systemd/system/gpdb.service

[root@cdw ~]# scp /etc/systemd/system/gpdb.service sdw2:/etc/systemd/system/gpdb.service

[root@cdw ~]# scp /etc/systemd/system/gpdb.service sdw3:/etc/systemd/system/gpdb.service

[root@cdw ~]# scp /etc/systemd/system/gpdb.service sdw4:/etc/systemd/system/gpdb.service

[root@cdw ~]#

########### 모든 노드에 gpdb.service 활성화

[root@cdw ~]# all

=> systemctl daemon-reload

=> systemctl enable gpdb.service

=> systemctl status gpdb.service

=> sync;sync;

########### 모든 노드 Reboot

########### 리부팅 후 cgroup v2 활성화 확인

[root@cdw ~]# all

=> ls /sys/fs/cgroup/gpdb.service

[ cdw.gpdbkr.com] cgroup.controllers cpuset.mems memory.max

[ cdw.gpdbkr.com] cgroup.events cpuset.mems.effective memory.min

[ cdw.gpdbkr.com] cgroup.freeze cpu.stat memory.numa_stat

[ cdw.gpdbkr.com] cgroup.max.depth cpu.weight memory.oom.group

[ cdw.gpdbkr.com] cgroup.max.descendants cpu.weight.nice memory.pressure

[ cdw.gpdbkr.com] cgroup.procs io.bfq.weight memory.stat

[ cdw.gpdbkr.com] cgroup.stat io.latency memory.swap.current

[ cdw.gpdbkr.com] cgroup.subtree_control io.max memory.swap.events

[ cdw.gpdbkr.com] cgroup.threads io.pressure memory.swap.high

[ cdw.gpdbkr.com] cgroup.type io.stat memory.swap.max

[ cdw.gpdbkr.com] cpu.max memory.current pids.current

[ cdw.gpdbkr.com] cpu.pressure memory.events pids.events

[ cdw.gpdbkr.com] cpuset.cpus memory.events.local pids.max

[ cdw.gpdbkr.com] cpuset.cpus.effective memory.high

[ cdw.gpdbkr.com] cpuset.cpus.partition memory.low

...

[sdw4.gpdbkr.com] cgroup.controllers cpuset.mems memory.max

[sdw4.gpdbkr.com] cgroup.events cpuset.mems.effective memory.min

[sdw4.gpdbkr.com] cgroup.freeze cpu.stat memory.numa_stat

[sdw4.gpdbkr.com] cgroup.max.depth cpu.weight memory.oom.group

[sdw4.gpdbkr.com] cgroup.max.descendants cpu.weight.nice memory.pressure

[sdw4.gpdbkr.com] cgroup.procs io.bfq.weight memory.stat

[sdw4.gpdbkr.com] cgroup.stat io.latency memory.swap.current

[sdw4.gpdbkr.com] cgroup.subtree_control io.max memory.swap.events

[sdw4.gpdbkr.com] cgroup.threads io.pressure memory.swap.high

[sdw4.gpdbkr.com] cgroup.type io.stat memory.swap.max

[sdw4.gpdbkr.com] cpu.max memory.current pids.current

[sdw4.gpdbkr.com] cpu.pressure memory.events pids.events

[sdw4.gpdbkr.com] cpuset.cpus memory.events.local pids.max

[sdw4.gpdbkr.com] cpuset.cpus.effective memory.high

[sdw4.gpdbkr.com] cpuset.cpus.partition memory.low

=> ls /sys/fs/cgroup/gpdb | wc -l

[sdw2.gpdbkr.com] 43

[scdw.gpdbkr.com] 43

[sdw3.gpdbkr.com] 43

[sdw1.gpdbkr.com] 43

[sdw4.gpdbkr.com] 43

[ cdw.gpdbkr.com] 43

=>

########### Greenplum cgroup v2 설정

[gpadmin@cdw ~]$ gpconfig -c gp_resource_manager -v "group-v2"

20240709:10:05:25:015578 gpconfig:cdw:gpadmin-[INFO]:-completed successfully with parameters '-c gp_resource_manager -v group-v2'

[gpadmin@cdw ~]$ gpstop -af

[gpadmin@cdw ~]$ gpstart -a

3. Disk IO 성능 제약 설정 테스트 결과

1) VM에서 Disk IO를 제약하지 않았을 경우

gpadmin=# ALTER RESOURCE GROUP rgoltp SET IO_LIMIT 'pg_default:wbps=2000,wiops=2000,rbps=2000,riops=2000';

[gpadmin@cdw gpkrtpch]$ ./2.2_upload.sh

./2.2_upload.sh: START TIME : 2024-03-28 21:42:52

./2.2_upload.sh: End TIME : 2024-03-28 21:43:31

./2.2_upload.sh|2024-03-28 21:42:52|2024-03-28 21:43:31|39 <<<==============

[gpadmin@cdw gpkrtpch]$

----system---- ----total-usage---- -dsk/total- -net/total- ------memory-usage-----

date/time |usr sys idl wai stl| read writ| recv send| used buff cach free

[sdw1] 28-03 21:43:11| 50 33 5 0 0| 27M 22M| 12M 5463k| 886M 34M 4096B 1335M

[sdw2] 28-03 19:59:25| 49 31 5 2 0| 26M 10M| 12M 5786k| 883M 26M 4096B 1356M

[sdw3] 28-03 19:59:24| 50 36 1 3 0| 41M 17M| 11M 5172k| 882M 35M 76k 1337M

[sdw4] 28-03 21:43:11| 52 29 5 0 0| 41M 22M| 11M 5267k| 889M 29M 4096B 1342M

2) VM에서 Disk IO를 제약할 경우

gpadmin=# ALTER RESOURCE GROUP rgoltp SET IO_LIMIT 'pg_default:wbps=2,wiops=2,rbps=2,riops=2';

[gpadmin@cdw gpkrtpch]$ ./2.2_upload.sh

./2.2_upload.sh: START TIME : 2024-03-28 21:36:24

./2.2_upload.sh: End TIME : 2024-03-28 21:42:05

./2.2_upload.sh|2024-03-28 21:36:24|2024-03-28 21:42:05|341 <<<==============

---------------------------------- Segment Node ----------------------------------

----system---- ----total-usage---- -dsk/total- -net/total- ------memory-usage-----

date/time |usr sys idl wai stl| read writ| recv send| used buff cach free

[sdw1] 28-03 21:41:30| 7 5 69 16 0|4699k 3514k|1767k 769k| 899M 24M 4096B 1308M

[sdw2] 28-03 19:57:44| 7 3 0 87 0|3197k 2810k|1587k 736k| 896M 19M 4096B 1317M

[sdw3] 28-03 19:57:44| 6 5 77 9 0|2717k 1565k|1712k 811k| 896M 28M 244k 1308M

[sdw4] 28-03 21:41:30| 9 5 68 15 0|1440k 2784k|2038k 984k| 901M 30M 4096B 1301M