oltpcore − Perform one run using the oltp workload
[-H] [-g|-G] [-a] [-n processes] [-k key] [-r runperiod] [-b] [-R runfile] [-W] [-K nn] [-p file] [-s file] [-o file] komment text ...
Execute one run using the oltp workload, saving intermediate results in the results directory and partially making final results available in the awr directory for browsing. Note that stdout and stderr are not saved.
Any arguments are added as extra komment text using -K option to rwloadsim; it must be preceded by -- if it includes any text that appear like options.
Options can be provided using gnu style long options as shown below. If a short options takes an argument, so does a long option.
-H|--help
Print short help.
-n|--processcount N
Set the number of processes, default 1
-k|--key key
Set the key that will be used when storing results in the repository database. The default is the same as the project name.
-r|--runperiod N
Set the runperiod in seconds; the default is 295s. You will get the nicest graphs if the value chosen is a bit less than a multiple of 300s. If less than 120s, no awr reports will be created.
-b|--simulatebatch
In stead of the normal simulation of an average arrival rate, run everything in a busy loop without any waiting/queuing taking place. You typically do this with much fewer threads (and/or processes) as each thread in each process constantly will be executing emulated business transactions.
Note that all burst features are turned off and that only the processing in the normal pool (set via pool_type in your project file) is done. You should normally arrange for your connections to be dedicated and set pool_type:="dedicated", but if you do use e.g. sessionpool, sessions will constantly be acquired and released.
-g|--graphs|-G|--qegraphs
Show graphs during the run using gnuplot’s output to X-Windows. This requires a properly set DISPLAY environment variable. The two latter options causes an alternative graph; see oltprun(2rwl).
-p|--runnumberfile file
Put the runnumber into file (default runnumber.txt)
-s|--startseconds file
Put the startseconds into file (default startseconds.txt)
-R|--runfile file
In stead of using the normal run.rwl file count in the oltp directory, use the file named. The primary use of this is for experiments that require modifications to run.rwl.
-a|--preallocate
If your orders and order_items tables are partitioned, you can use the -a option to make sure an empty set of partitions are created at the start of the run. Only use this if the automatic allocation via interval partitions appears to cause trouble
-o|--extraoutput file
Add the contents of the named text file to generated html output; this can be used if you want to add more information than feasible using komment.
-i|--integer|-d|--double spec
Provide the option to rwloadsim when doing the actual run, i.e. when run.rwl (or the file given as the -R option) is executed.
-W|--awrwait
-K|--awrkill n
See below under COMPLETION
Under normal circumstances, all the processes started by oltpcore should complete, and it should therefore be feasible to simple have a "wait" at the end of the script. If, however, one or more of the processes take an unexpectedly long time to complete or even never completes, such a "wait" would also not complete (in reasonable time). Therefore, oltpcore has a mechanism that will kill processes taking longer than expected. For most processes that do actual simulation work, a consequence of this is simply that the process gets canceled, but for the process gathering awr and other system statistics, it means no awr will be produced. Oltpcore therefore waits as long as possible until killing the awr gathering process, but it will do so eventually. If you find that your awr reports does not get created even when things otherwise are running normally, you can change the behavior in one of two ways:
By using the -W option, oltpcore will always wait for the awr gathering process which does increase the risk that some system failure will cause oltpcore to wait for a very long time or even not complete.
By using the -K nn option, you can change the default time of 15s before the final kill will be performed. The implication is that if there are processes to be killed, the final wait time (and therefore total execution time) will be longer, but oltpcore is guaranteed to complete.
The oltpcore command shows stdout and stderr to the terminal, so you can immediately see any errors, but only stdout will be save to a file. You should therefore only use this command initially while you verify everything is working as expected. Otherwise, always use the oltprun(2rwl) command that makes both stdout and stderr available for later browsing, but only outputs stdout to the terminal.
Copyright
© 2023 Oracle Corporation
Licensed under the Universal Permissive License v 1.0 as
shown at https://oss.oracle.com/licenses/upl
oltpsetup(2rwl), oltprun(2rwl)