Request logging logs individual requests received from, and responses sent to, an application. It is most useful for determining what the database server is being asked to do by the application.
Request logging is also a good starting point for performance analysis of a specific application when it is not obvious whether the database server or the client is at fault. You can use request logging to determine the specific request to the database server that might be responsible for problems.
All the functionality and data provided by the request logging feature is also available using diagnostic tracing. Diagnostic tracing also offers additional features and data.
Logged information includes such things as timestamps, connection IDs, and request type. For queries, it also includes the isolation level, number of rows fetched, and cursor type. For INSERT, UPDATE, and DELETE statements, it also includes the number of rows affected and number of triggers fired.
The request log contains all statements with obfuscated sensitive information. The only case when sensitive information is not obfuscated is a statement with a parsing error.
You can use the -zr server option to turn on request logging when you start the database server. You can redirect the output to a request log file for further analysis using the -zo server option. The -zn and -zs option let you specify the number of request log files that are saved and the maximum size of request log files.
For more information about these options, see:
These server options do not impact diagnostic tracing in Sybase Central. File-based request logging is completely separate from the diagnostic tracing feature in Sybase Central, which makes use of dbo-owned diagnostic tables in the database to store request log information.
The sa_get_request_times system procedure reads a request log and populates a global temporary table (satmp_request_time) with statements from the log and their execution times. For INSERT/UPDATE/DELETE statements, the time recorded is the time when the statements were executed. For queries, the time recorded is the total elapsed time from PREPARE to DROP (describe/open/fetch/close). That means you need to be aware of any open cursors.
Analyze satmp_request_time for statements that could be candidates for improvements. Statements that are inexpensive, but frequently executed, may represent performance problems.
You can use sa_get_request_profile to call sa_get_request_times and summarize satmp_request_time into another global temporary table called satmp_request_profile. This procedure also groups statements together and provides the number of calls, execution times, and so on.
If the log is being analyzed using the tracetime.pl Perl script, the max_client_statements_cached option should be set to 0 to disable client statement caching while the request log is captured.
Discuter à propos de cette page dans DocCommentXchange.
|Copyright © 2013, SAP AG ou société affiliée SAP - SAP Sybase SQL Anywhere 16.0|