This subsection discusses problems that might occur during system operation and how to handle them.
The following describes the action to be taken when data cannot be sent or received between the stream data processing engine and adaptors:
If an overflow has occurred in the input stream queue for the stream data processing engine, control the volume of input adaptor processing. By changing the number of records that are read at one time or the interval at which records are sent to the SDP server, you can reduce the amount of memory used per unit of time.
You can change the number of records by using the readUnit attribute of the input tag in file input connector definition in the adaptor configuration definition file.
You can change the transmission interval by using the interval attribute in the CB definition for input.
If an interval conflict message is output for a query group or the SDP server, collect data by referencing 6.2 Data to be collected in the event of an error, and then contact the customer engineer.
If a failure occurs while a query group is executing, Stream Data Platform - AF shuts down the query group. If a query group is shut down, the tuples accumulated in the stream queues are discarded.
The following describes how to handle query group shutdown for each cause.
KFSP42005-E The upper limit value of the stream queue was exceeded. Stream name = ..., number of elements = ..., upper limit value = ... |
KFSP420**-E =... |
KFSP42301-E The number of tuples held by the timestamp adjustment function exceeded the upper limit value.... |
If query processing required too much time and a timeout resulted, see 6.2 Data to be collected in the event of an error, collect the required data, and determine the cause of the error. If necessary, take appropriate action, such as forcing termination of the SDP server.
For details about forced termination of the SDP server, see sdpstop (stops the SDP server) in 7. Commands.
If adaptor processing has slowed, check to see if full garbage collections have been occurring frequently. All adaptor processing stops while a full garbage collection is underway, resulting in a slowdown of processing speed. You can obtain information about full garbage collection from the following file, as applicable:
working-directory\logs\SDPServerVMinteger.log
working-directory\logs\SDPClientVMinteger.log
Note that integer in the file name is a number in the range from 01 to 04 (because a maximum of four files can be used). The output destination and names of these files are set to the default values.
The following shows an example of output of a full garbage collection:
[VGC]<Fri Jan 08 17:19:46.159 2010>[Full GC 54330K->45174K(519296K), |
If the underlined Full GC is output frequently or garbage collection takes several seconds, JavaVM assumes that not enough memory is allocated to the heap area. In such a case, consider taking one of the corrective actions described below:
If the data type specified in the type attribute of the field tag in the format conversion definition in the adaptor configuration definition file does not match the data type in the corresponding CQL, an error occurs during tuple transmission and the adaptor processing stops. In such a case, check and revise (if necessary) the data type.
For details about the type attribute of the field tag in the format conversion definition, see 9.11.1 Format conversion definition.
If all tuple log buffers become full, some tuple logs will be lost.
If tuple logs have been lost, the result of query re-execution by the sdptplput command might not match the result obtained by executing queries using adaptors. Therefore, you must specify the appropriate number and sizes of buffers so that a sufficient number of tuples can be acquired.
The formula for estimating the tuple log buffers is shown below. Using this formula, determine the number and sizes of tuple log buffers that you need for storing tuples.
Size of one tuple log buffer (bytes) = size of one tuple x number of tuples to be stored |
For the size of one tuple, see 2.7.1(2) Memory required for one tuple.
If the analysis results cannot be displayed on the dashboard, check the following:
If an error occurs on the dashboard, an error dialog box is displayed in the dashboard window. This dialog box displays the following information:
Displayed name | Description |
---|---|
"ErrID" | Displays an error message ID |
"ErrMsg" | Displays the error message |
For details about the messages, see the manual uCosminexus Stream Data Platform - Application Framework Messages.