Applicable versions and levels
Solaris versions: 5.0 or later
Linux versions: 5.2, V10.0L10 or later
Linux for Itanium versions: V12.0L10 or later
HP-UX versions: 5.1 or later
Action
Cause
If many job nets and jobs are executed at the same time, and there are many message startup job nets, the startup of the job nets and jobs may be delayed, and it may not be possible to correctly recognize the Job Execution Control attribute job status using Jobscheduler.
Action method
Perform the following procedure to extend the message queue:
Solaris 7/8/9:
Add the following settings to the /etc/system file, and then restart the system after editing the file.
set msgsys:msginfo_msgtql = <number of job nets to be started at the same time> (*1) x 4 |
*1: To use a Job Execution Control attribute and an Interstage attribute job net, also add the number of jobs to be started at the same time.
*2: "200" is the standard value, however, this will vary depending on the length of the path for the host name or output file. Perform sufficient verification, and set an appropriate value.
*3: In Solaris8 or later, there is no need to set the "msgsys:msginfo_msgseg" and "msgsys:msginfo_msgmap" values.
*4: Request the msgsys:msginfo_msgtql value first, and then calculate using that value as well.
*5: This is for EE versions. In SE versions, set the number of subsystems used as "1".
Remarks: msgsys:msginfo_msgmnb and msgsys:msginfo_msgmap are parameters that specify the maximum value, therefore correct the current value if it is less than the above values. For other parameters, add the above values to the current values.
Solaris 10
Change the values set in the /etc/project file to the following values:
process.max-msg-qbytes=(privileged,<number of job nets to be started at the same time> (*1) x 200 (*2),deny) |
*1: Additionally, if using a job with Job Execution Control attributes, add the number of jobs to be started simultaneously.
*2: "200" is the standard value, however, this will vary depending on the length of the path for the host name or output file. Perform sufficient verification, and set an appropriate value.
*3: This is for EE versions. In SE versions, set the number of subsystems used as "1".
Remarks: The max-msg-qbytes and max-msg-ids operating system default values are as follows:
max-msg-qbytes: 64KB
max-msg-ids: 128
HP-UX versions:
Use a system manager to change the kernel parameters, and then create the kernel again.
msgtql <number of job nets to be started at the same time> (*1) x 4 |
*1: Additionally, if using a job with Job Execution Control attributes, add the number of jobs to be started simultaneously.
*2: "200" is the standard value, however, this will vary depending on the length of the path for the host name or output file. Perform sufficient verification, and set an appropriate value.
*3: Request the msgtql value first, and then calculate based on that value.
*4: This is for EE versions. In SE versions, set the number of subsystems used as "1".
Remarks: msgmnb and msgmap are parameters that specify the maximum value, therefore correct the current value if it is less than the above values. For other parameters, add the above values to the current values.
Linux versions:
Add the following settings to the /etc/sysctl.conf file, and then restart the system after editing the file.
kernel.msgmnb = <number of job nets to be started at the same time> (*1) x 200 (*2) |
*1: Additionally, if using a job with Job Execution Control attributes, add the number of jobs to be started simultaneously.
*2: "200" is the default value, however, this will vary depending on the length of the path for the host name or output file. Perform sufficient verification, and set an appropriate value.
*3: This is for EE versions. In SE versions, set the number of subsystems used as "1".
Remarks: kernel.msgmnb is a parameter that specifies the maximum value, therefore correct the current value if it is less than the above values. For other parameters, add the above values to the current values.
(EE versions only)
To perform multi-subsystem operations in Linux versions, the number of message queues used will be four times greater than the number of subsystems. For this reason, the maximum number of message queues that can be used may be exceeded, which will mean that the subsystem cannot be started.
In this case, add the settings shown below to the /etc/sysctl.conf file to extend the maximum number of message queues that can be used in the system. After editing the file, restart the system.
kernel.msgmni (maximum number of message queues that can be used) = 4 x <number of subsystems to be used> |
