I documented the fields and their interpretation in another location. This blog post is my views on the data and what you can do with it.
You can control how often the data is published. See here from IBM.
I ran a workload putting on and getting a non persistent message to a queue for a minute.
- My laptop has 4 CPUs and was running Ubuntu 18.04
- My application reported it used 42 seconds of CPU in 60 seconds elapsed time
- CPUSTAT tool reported my application using 70% of a CPU, and /opt/mqm/bin/amqzlaa0 running total 43% usr 30% sys 13%
- MQ reported Total 10.7% Usr 7.4 % sys 3.3
It looks like CPUSTAT reports usage of one CPU, and MQ reports it as %usage of all the CPUs.
If the number of CPUs is changed – then the MQ value will be different.
I do not think the data reported by MQ is very useful. If the workload was moved to a differently configured machine, for example a different number of CPU or faster CPUs, the %CPU may change – but you do not know why it has changed.
If this number is trending upwards you may want to look at the workload and do capacity planning.
Rerunning the same test and looking at different MQ metrics
- User CPU time percentage 19.01% this looks like 19 % of 4 CPUs
- System CPU time percentage 9.66% this looks like 19 % of 4 CPUs
- CPU load – one minute average was 1.27. This looks like 1.27 CPUs worth of CPU were being used. CPUSTAT reported 70% (user+sys) for my application + 43% for MQ,which is 113 which is close enough. It looks like CPU load – one minute average is reporting usage of one CPU.
Publication received PutDate:20190303 PutTime:18201013 Interval:10.000 seconds
Log – bytes in use 50331648
Log – bytes max 83886080
Log file system – bytes in use 16434135040
Log file system – bytes max 19549782016
Log – physical bytes written 33718272 3371827/sec
Log – logical bytes written 5418954 541895/sec
Log – write latency 994 uSec
Log – write size 4099
Log – current primary space in use 18.79%
Log – workload primary space utilization 53.33%
The figures I think are useful are in a bold font.
Seehere for my blog entry on using the log data, and what the fields mean.
From a high level view point – I think the absolute values of, and rates of
Class 2 Type 3: STATMQI/PUT
- Interval total MQPUT/MQPUT1 count
- Interval total MQPUT/MQPUT1 byte count
- Put non-persistent messages – byte count
- Put persistent messages – byte count
Are useful metrics for a high level view of the status of the queue manager