C.1.5 Core Dump of a Kernel Zone

In Kernel Zones environment, a forcible stop is performed by both the KZONE Panic shutdown agent and the KZONE Reset shutdown agent. A crush dump is output by only the KZONE Panic shutdown agent.

However, by setting a core dump output as necessary, a core dump of a Kernel Zone can be collected by the KZONE Reset shutdown agent.

This section explains how to set a core dump output.

Note

Setting a core dump output is optional.
If the core dump output is set, the KZONE Reset shutdown agent waits until the collection of the core dump, and then it stops the Kernel Zone forcibly. Therefore, the required time for a failover that causes a node error and a double fault of a resource may be long.
The required time for outputting a core dump and for executing the halt subcommand of the zoneadm command depend on the I/O performance of a disk and memory usage.
Therefore, calculate the time after completing a construction of the cluster application, as well as letting the cluster application operated.

For how to set the core dump output in the Kernel Zone, perform the following procedure on all nodes which constitute a cluster.

Stopping the shutdown facility
Execute the following command to stop the shutdown facility.
```
# /opt/SMAW/SMAWsf/bin/sdtool -e
```

Changing the timeout value for the KZONE Reset shutdown agent (SA_kzoner)

Edit the rcsd.cfg file using an editor such as vi.

# vi /etc/opt/SMAW/SMAWsf/rcsd.cfg

CFNameX,weight=weight,admIP=myadmIP: agent=SA_kzonep,timeout=p_timeout:agent=SA_kzoner,timeout=r_timeout:agent=SA_kzchkhost,timeout=c_timeout
CFNameX,weight=weight,admIP=myadmIP: agent=SA_kzonep,timeout=p_timeout:agent=SA_kzoner,timeout=r_timeout:agent=SA_kzchkhost,timeout=c_timeout

r_timeout: Specifies a timeout period (in seconds) for the KZONE Reset shutdown agent.

Calculate the value using the following calculation formula.

r_timeout = Required time for creating a file with the memory size set in the Kernel Zone + 
            Required time for executing the halt subcommand of the zoneadm command + 20 seconds

* The minimum value of the r_timeout value is default 70 seconds. If the calculation result for the r_timeout value is less than 70 seconds, use the default value of 70 seconds.

Example

When changing the timeout period for the KZONE Reset shutdown agent from 70 seconds to 100 seconds

Before edit

CFNameX,weight=weight,admIP=myadmIP: agent=SA_kzonep,timeout=45:agent=SA_kzoner,timeout=70:agent=SA_kzchkhost,timeout=20
CFNameX,weight=weight,admIP=myadmIP: agent=SA_kzonep,timeout=45:agent=SA_kzoner,timeout=70:agent=SA_kzchkhost,timeout=20

After edit

CFNameX,weight=weight,admIP=myadmIP: agent=SA_kzonep,timeout=45:agent=SA_kzoner,timeout=100:agent=SA_kzchkhost,timeout=20
CFNameX,weight=weight,admIP=myadmIP: agent=SA_kzonep,timeout=45:agent=SA_kzoner,timeout=100:agent=SA_kzchkhost,timeout=20

Calculate each processing time on the global zone host where the Kernel Zone operates is as follows.

In the following example, the timeout value is 70 + 10 + 20 = 100 seconds.

Example: The required time for creating a file with the memory size set in the Kernel Zone and for executing the halt subcommand of the zoneadm command when the memory size of the Kernel Zone kzone1 is 8 GB.

# /usr/bin/time mkfile 8G /var/crash/tmpfile
real       70.0
user        0.0
sys         0.0
# /usr/bin/time zoneadm -z kzone1 halt
real       10.0
user        0.0
sys         0.0
#

The memory size for the Kernel Zone can be checked with the zonecfg command as follows.

# zonecfg -z kzone1 info
zonename: kzone1
brand: solaris-kz
...
capped-memory:
     physical: 8G
...
#

Creating a file for enabling the core dump collection function
Execute the following command to enable the core dump collection function.
```
# touch /etc/opt/SMAW/SMAWsf/enablekzcore.cfg
# chmod 600 /etc/opt/SMAW/SMAWsf/enablekzcore.cfg
```
* The core dump is output to the following directory on the global zone host.
```
/var/crash/<Kernel Zone name>.<Execution time for the savecore subcommand(YYYYMMDDHHMMSS format)>.kzcore
```
Example: When the core dump of kzone1 was output at 12:34:56 January 1st, 2015
```
/var/crash/kzone1.20150101123456.kzcore
```
Starting the shutdown facility
Execute the following command to start the shutdown facility.
```
# /opt/SMAW/SMAWsf/bin/sdtool -b
```