Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Oracle Solaris 11 Advanced Administration Cookbook.docx
Скачиваний:
3
Добавлен:
01.07.2025
Размер:
3 Мб
Скачать
☆

Value is basic (only the owner can modify it).

f deny: This can deny any requested resource for an amount above the limit value (500 MB). The other possibilities would be none (no action is taken even if the requested resource is above the limit) and signal, in which a signal is sent when the threshold value is exceeded.

Resource capping is a service implemented by the rcapd daemon, and this service can be enabled by the following command:

root@solaris11-1:~# svcs -a | grep rcap

disabled 21:56:20 svc:/system/rcap:default

root@solaris11-1:~# svcs -d svc:/system/rcap:default

STATE STIME FMRI

online 21:56:31 svc:/system/filesystem/minimal:default

online 21:56:33 svc:/system/resource-mgmt:default

online 21:56:35 svc:/system/manifest-import:default root@solaris11-1:~# svcadm enable svc:/system/rcap:default root@solaris11-1:~# svcs - a | grep rcap

online 22:52:06 svc:/system/rcap:default root@solaris11-1:~# svcs -p svc:/system/rcap:default STATE STIME FMRI

online 22:52:06 svc:/system/rcap:default

22:52:06 5849 Rcapd

Chapter 4

Reboot zone1 for memory capping to take effect. It would be feasible to enable the resource capping daemon without rebooting and starting the daemon now by running the following command:

root@solaris11-1:~# rcapadm -E -n

To monitor the action of the rcap daemon (rcapd), execute the following commands:

root@solaris11-1:~# zoneadm -z zone1 shutdown -r

root@solaris11-1:~# zoneadm list -iv

ID NAME STATUS PATH BRAND IP

0 global running / solaris shared

1

zone2

running

/myzones/zone2

solaris

excl

3

zone1

running

/myzones/zone1

solaris

excl

root@solaris11-1:~# rcapstat -z

id

zone

nproc vm

rss

cap

at

avgat

pg

avgpg

3

zone1

- 26M

38M

1024M

0K

0K

0K

0K

3

zone1

- 31M

44M

1024M

0K

0K

0K

0K

3

zone1

- 31M

44M

1024M

0K

0K

0K

0K

The used physical memory (RSS) is below the memory capping limit (1024 MB). If the physical memory is increased, its limit is 1024 MB. Nice!

To make this example more attractive, some changes can be made. Let's remove the first_pool resource pool (and any other existing pool) from zone1. Additionally, the first_pool pool will be deleted by the pooladm -x command. Obviously, the new pool configuration must be saved by the pooladm -s command. The following is the sequence:

root@solaris11-1:~# zonecfg -z zone1 clear pool root@solaris11-1:~# zoneadm -z zone1 shutdown -r root@solaris11-1:~# pooladm -x

root@solaris11-1:~# pooladm -s root@solaris11-1:~# pooladm system default

string system.comment int system.version 1

boolean system.bind-default true

string system.poold.objectives wt-load

pool pool_default

int pool.sys_id 0

Zones

boolean pool.active true boolean pool.default true int pool.importance 1 string pool.comment

pset pset_default

pset pset_default

int pset.sys_id -1 boolean pset.default true uint pset.min 1

uint pset.max 65536

string pset.units population uint pset.load 15511

uint pset.size 4 string pset.comment

cpu

int cpu.sys_id 1 string cpu.comment

string cpu.status on-line

cpu

int cpu.sys_id 3 string cpu.comment

string cpu.status on-line

cpu

int cpu.sys_id 0 string cpu.comment

string cpu.status on-line

cpu

int cpu.sys_id 2 string cpu.comment

string cpu.status on-line

Chapter 4

Everything has returned to the default status, and from this point, zone1 doesn't have a special associated pool. This permits us to focus on FSS from now on.

The following command checks what the current default kernel scheduling class is:

root@solaris11-1:~# dispadmin -d

dispadmin: Default scheduling class is not set

There is no default scheduling class. If we want to use FSS, then it would be appropriate to configure it on the global zone because this setting will be inherited by all non-global zones. To configure the FSS as explained, execute the following command:

root@solaris11-1:~# dispadmin -d FSS

root@solaris11-1:~# dispadmin -d

FSS (Fair Share)

This setup only takes effect after a system is rebooted. After the system has been reinitiated, all processes will be classified as FSS. Nonetheless, to enforce it now without a reboot, execute the following command:

root@solaris11-1:~# priocntl -s -c FSS

Unfortunately, all current processes are still running under other scheduling classes and only new processes will take the FSS setting. This can be verified by running the following command:

root@solaris11-1:~# ps -efcZ | more

ZONE

UID

PID

PPID

CLS

PRI

STIME

TTY

TIME

CMD

global

root

0

0

SYS

96

00:04:41

?

0:01

sched

global

root

5

0

SDC

99

00:04:38

?

0:07

zpool-rpool

global

root

6

0

SDC

99

00:04:42

?

0:01

kmem_task

global

root

1

0

TS

59

00:04:42

?

0:00

/usr/sbin/init

global

root

2

0

SYS

98

00:04:42

?

0:00

pageout

global

root

3

0

SYS

60

00:04:42

?

0:00

fsflush

global

root

7

0

SYS

60

00:04:42

?

0:00

intrd

global

root

8

0

SYS

60

00:04:42

?

0:00

vmtasks

global

root

115

1

TS

59

00:05:09

?

0:00

/usr/lib/pfexecd

global svc.startd

root

11

1

TS

59

00:04:48

?

0:13

/lib/svc/bin/

global root

13

1

TS

59

00:04:48

?

0:33

/lib/svc/bin/

svc.configd

global

root

911

1

TS

59

02:05:55

?

0:00

(truncated output)

Zones

Again, it's unnecessary to wait for the next reboot. Therefore, all processes can be moved from their current scheduling classes to FSS by executing the following commands:

root@solaris11-1:~# priocntl -s -c FSS -i all

root@solaris11-1:~# ps -efcZ | more

ZONE

UID

PID

PPID

CLS

PRI

STIME

TTY

TIME

CMD

global

root

0

0

SYS

96

00:04:41

?

0:01

sched

global

root

5

0

SDC

99

00:04:38

?

0:12

zpool-rpool

global

root

6

0

SDC

99

00:04:42

?

0:02

kmem_task

global

root

1

0

FSS

29

00:04:42

?

0:00

/usr/sbin/init

global

root

2

0

SYS

98

00:04:42

?

0:00

pageout

global

root

3

0

SYS

60

00:04:42

?

0:01

fsflush

global

root

7

0

SYS

60

00:04:42

?

0:00

intrd

global

root

8

0

SYS

60

00:04:42

?

0:00

vmtasks

global pfexecd

root

115

1

FSS

29

00:05:09

?

0:00

/usr/lib/

global svc.startd

root

11

1

FSS

29

00:04:48

?

0:13

/lib/svc/bin/

global svc.configd

root

13

1

FSS

29

00:04:48

?

0:33

/lib/svc/bin/

(truncated

output)

When FSS is set up as the default scheduling class in the global zone, all non-global zones automatically take this configuration. To verify this, run the following command:

root@solaris11-1:~# zlogin zone1

[Connected to zone 'zone1' pts/4]

Oracle Corporation SunOS 5.11 11.1 September 2012 root@zone1:~# ps -efcZ | more

ZONE UID PID PPID CLS PRI STIME TTY TIME CMD

zone1 root 3944 2454 FSS 29 02:06:47 ? 0:00 /usr/sbin/ init

zone1 root 4284 2454 FSS 29 02:06:58 ? 0:06 /lib/svc/ bin/svc.startd

zone1

root

2454

2454

SYS

60

02:06:29

?

0:00

zsched

zone1

root

5479

2454

FSS

59

02:48:52

pts/4

0:00

/usr/bin/

login -z global

-f root

zone1 root bin/svc.configd

4287

2454

FSS

29

02:07:00

?

0:21

/lib/svc/

zone1 netcfg netcfgd

4448

2454

FSS

29

02:07:27

?

0:00

/lib/inet/

zone1 root

4922

2454

FSS

29

02:08:21

?

0:00

(truncated output)

Chapter 4

We can realize that all main processes from zone1 are under the FSS class. Anyway, it is recommended that the FSS class be explicitly configured in the non-global settings in order to prevent possible mistakes in the future. Therefore, execute the following command:

root@solaris11-1:~# zonecfg -z zone1 zonecfg:zone1> set scheduling-class=FSS zonecfg:zone1> verify

zonecfg:zone1> commit zonecfg:zone1> exit root@solaris11-1:~#

root@solaris11-1:~# zonecfg -z zone2 zonecfg:zone2> set scheduling-class=FSS zonecfg:zone2> verify

zonecfg:zone2> commit zonecfg:zone2> exit root@solaris11-1:~#

Finally, it is the right moment to use the FSS class to configure some shares for each zone (zone1 and zone2). This way, it is possible to share an amount (70 percent) from the CPU processing for zone1 and the other amount (30 percent) from the CPU processing for zone2. The following is the procedure:

root@solaris11-1:~# zonecfg -z zone1

zonecfg:zone1> add rctl

zonecfg:zone1:rctl> set name=zone.cpu-shares

zonecfg:zone1:rctl> add value (priv=privileged, limit=70,action=none)

zonecfg:zone1:rctl> end zonecfg:zone1> verify zonecfg:zone1> commit zonecfg:zone1> exit

root@solaris11-1:~# zonecfg -z zone2

zonecfg:zone2> add rctl

zonecfg:zone2:rctl> set name=zone.cpu-shares

zonecfg:zone2:rctl> add value (priv=privileged,limit=30,action=none)

zonecfg:zone2:rctl> end zonecfg:zone2> verify zonecfg:zone2> commit zonecfg:zone2> exit

Zones

This is excellent! Shares were assigned to zone1 (70 shares) and zone2 (30 shares) using the zonecfg command in a persistent way. For both the zones to take effect, execute the following commands:

root@solaris11-1:~# zoneadm -z zone1 shutdown -r

root@solaris11-1:~# zoneadm -z zone2 shutdown -r

The processor time can be followed and monitored using the following command:

root@solaris11-1:~# prstat -Z

PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP

4466 root

216M

98M

sleep

59

0

0:00:41

0.7%

java/25

4702 root

129M

19M

sleep

59

0

0:00:06

0.5%

gnome-

terminal/2 rcapd/1

5 root 0K 0K sleep 99 -20 0:00:19 0.2% zpool- rpool/138

898 root 53M 18M sleep 53 0 0:00:06 0.1% poold/9

(omitted output)

automountd/2

198

root

1780K

788K

sleep

29

0

0:00:00

0.0%

utmpd/1

945

root

2392K

1552K

sleep

59

0

0:00:00

0.0%

ttymon/1

ZONEID

NPROC

SWAP

RSS

MEMORY

TIME

CPU

ZONE

0

117

2885M

794M

9.5%

0:03:28

2.5%

global

2

28

230M

62M

0.7%

0:00:30

0.1%

zone1

1

28

230M

64M

0.7%

0:00:29

0.1%

zone2

Surprisingly, it is feasible to change the zone.cpu-shares attribute dynamically without rebooting zones but in a non-persistent way (all the changes are lost after a reboot) by running the following commands:

root@solaris11-1:~# prctl -n zone.cpu-shares -v 60 -r -i zone zone1

root@solaris11-1:~# prctl -n zone.cpu-shares -P -i zone zone1

zone: 3: zone1

zone.cpu-shares usage 60 - - - zone.cpu-shares privileged 60 - none -

zone.cpu-shares system 65535 max none -

root@solaris11-1:~# prctl -n zone.cpu-shares -v 40 -r -i zone zone2

root@solaris11-1:~# prctl -n zone.cpu-shares -P -i zone zone2

Chapter 4

zone: 4: zone2

zone.cpu-shares usage 40 - - - zone.cpu-shares privileged 40 - none -

zone.cpu-shares system 65535 max none - root@solaris11-1:~#

To collect information about the memory and CPU from both zones in an interval of 5 seconds, execute the following command:

root@solaris11-1:~# zonestat -z zone1,zone2 -r physical-memory 5

Collecting data for first interval... Interval: 1, Duration: 0:00:05

PHYSICAL-MEMORY SYSTEM MEMORY

mem_default 8191M

ZONE USED %USED CAP %CAP [total] 1464M 17.8% - -

[system] 624M 7.62% - - zone2 63.9M 0.78% - - zone1 3561K 0.04% 1024M 0.33%

Interval: 2, Duration: 0:00:10

PHYSICAL-MEMORY SYSTEM MEMORY

mem_default 8191M

ZONE USED %USED CAP %CAP [total] 1464M 17.8% - -

[system] 624M 7.62% - - zone2 63.9M 0.78% - - zone1 3485K 0.04% 1024M 0.33%

Removing all configured shares is quickly executed by running: root@solaris11-1:~# zonecfg -z zone1 clear cpu-shares root@solaris11-1:~# zonecfg -z zone2 clear cpu-shares root@solaris11-1:~# zoneadm -z zone1 shutdown -r root@solaris11-1:~# zoneadm -z zone2 shutdown -r

Zones

Keeping up with our approach about the resource manager, there's a zone resource,

named dedicated-cpu, where it is possible to specify a subset of processors (or cores)

to a non-global zone. For example, the following example shows us that zone1 can use one to four processors (ncpus=1-4) according to the demand, and this setting has an importance value equal to 8 when competing for resources against other zones or configurations. This smart setup creates a temporary pool including any necessary processor inside it. The following is the sequence:

root@solaris11-1:~# zonecfg -z zone1 zonecfg:zone1> add dedicated-cpu zonecfg:zone1:dedicated-cpu> set ncpus=1-4 zonecfg:zone1:dedicated-cpu> set importance=8 zonecfg:zone1:dedicated-cpu> end zonecfg:zone1> verify

zonecfg:zone1> commit

zonecfg:zone1> exit

root@solaris11-1:~# zoneadm -z zone1 shutdown -r

root@solaris11-1:~# zlogin zone1

[Connected to zone 'zone1' pts/2]

Oracle Corporation SunOS 5.11 11.1 September 2012 root@zone1:~# pooladm

system default

string system.comment int system.version 1

boolean system.bind-default true

string system.poold.objectives wt-load

pool SUNWtmp_zone1 int pool.sys_id 1

boolean pool.active true boolean pool.default false int pool.importance 8 string pool.comment boolean pool.temporary true pset SUNWtmp_zone1

pset SUNWtmp_zone1

int pset.sys_id 1

Chapter 4

boolean pset.default false uint pset.min 1

uint pset.max 4

string pset.units population uint pset.load 4

uint pset.size 2 string pset.comment

boolean pset.temporary true

cpu

int cpu.sys_id 1 string cpu.comment

string cpu.status on-line

cpu

int cpu.sys_id 0 string cpu.comment

string cpu.status on-line

Amazing! To remove the dedicated-cpu resource from zone1, execute the following command:

root@solaris11-1:~# zonecfg -z zone1 zonecfg:zone1> remove dedicated-cpu zonecfg:zone1> verify zonecfg:zone1> commit zonecfg:zone1> exit

Before continuing, we must reboot the zone by running the following command:

root@solaris11-1:~# zoneadm -z zone1 shutdown -r

Another good technique to control zone resources is using the capped-cpu resource, which permits us to specify how big a percentage of a CPU the zone can use. The value to be specified means a percentage of CPUs, and this procedure can be performed by executing the following sequence:

root@solaris11-1:~# zonecfg -z zone1 zonecfg:zone1> add capped-cpu zonecfg:zone1:capped-cpu> set ncpus=2.5 zonecfg:zone1:capped-cpu> end

Zones

zonecfg:zone1> verify zonecfg:zone1> commit zonecfg:zone1> exit

root@solaris11-1:~# zoneadm -z zone1 shutdown -r

According to the previous configuration, the ncpus=2.5 attribute means 250 percent of CPUs or 2.5 CPUs. To remove the recently added resource, execute the following command:

root@solaris11-1:~# zonecfg -z zone1 zonecfg:zone1> remove capped-cpu zonecfg:zone1:capped-cpu> end zonecfg:zone1> verify zonecfg:zone1> commit zonecfg:zone1> exit

After all the changes, we have to reboot the zone by executing the following command:

root@solaris11-1:~# zoneadm -z zone1 shutdown -r

This is outstanding! We have executed many trials with resource management, and all of them have worked! As zone1 still has a resource capping (memory), it is time to remove it:

root@solaris11-1:~# zonecfg -z zone1 zonecfg:zone1> remove capped-memory zonecfg:zone1> verify zonecfg:zone1> commit zonecfg:zone1> exit

root@solaris11-1:~# zoneadm -z zone1 shutdown -r

Finally, the resource capping feature can be disabled by executing the following command:

root@solaris11-1:~# svcs -a | grep rcap

online 18:49:28 svc:/system/rcap:default root@solaris11-1:~# rcapadm -D

state: disabled memory cap enforcement threshold: 0%

process scan rate (sec): 15 reconfiguration rate (sec): 60

report rate (sec): 5 RSS sampling rate (sec): 5

root@solaris11-1:~# svcs -a | grep rcap

disabled 19:28:33 svc:/system/rcap:default

Chapter 4

Another way of disabling the resource capping feature would be to execute the following command:

root@solaris11-1:~# svcadm disable svc:/system/rcap:default

Perfect! Everything has returned to the initial setup.

An overview of the recipe

This section was very long, and we could learn lots of details about resource management controls and how to limit processors and the memory. In the next chapter, we are going to handle the network resource control.

Соседние файлы в предмете [НЕСОРТИРОВАННОЕ]