Browse Prior Art Database

A Method and Apparatus for Asynchronously Log Processing in Application Servers

IP.com Disclosure Number: IPCOM000210664D
Publication Date: 2011-Sep-08
Document File: 4 page(s) / 122K

Publishing Venue

The IP.com Prior Art Database

Abstract

During application stress, if some trace or debug is enabled to diagnose, sometimes the stress issues can't be reproduced as a result of the heavy I/O. A local log server is defined. When a business server encounters a log/trace/debug writing action, it just sends the message to the local log server and continues with its following code.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 4

A Method and Apparatus for Asynchronously Log Processing in Application Servers

In some customer cases and SVT work, we often meet a problem in the following pieces of work:

1) During the application stress run, we find some issues happening. For example, exceptions/errors are reported or business transactions fail;
2) Some trace or debug information is enabled to diagnose the issues;
3) The issues can't be reproduced with the trace/debug enabled.

So the problem is that some stress issues can't be reproduced with trace/debug enabled.

This problem is caused by:

In application servers, ancillary data like log/trace is created and written to output systems within the current running logic. For example,

(This page contains 00 pictures or other non-text object)

The log actions usually write messages to files. These I/O operations are very time consuming and will reduce the execution frequency during workload stress as the business logics need to waitfor the log I/O to complete. As we know, some issues only happen during the stress with high execution frequency. So if trace/debug is enabled during stress for diagnosing, it will be hard or impossible to reproduce such kind of issues. This problem often happens in customer cases as well as stress test in SVT.

Currently we can only narrow down the trace/debug specifications which will enable less trace/debug to reduce the I/O affect as much as possible. But this will make the problem determination more difficult as the trace/debug might be not quite enough.

How to reduce the trace/debug I/O affect on stress is a valuable problem within application server problem determination area. We searched in google patent with key words "asynchronouslog" and found few methods for it. We will take the following one for example:

1


Page 02 of 4

Method and apparatus for asynchronous log message processing
(US 20070234287A1)

It provides a method to uncouple the log messages produced within a database transaction. It can save the log of a failed database transaction without saving the work itself.

In our opinion, it has the following disadvantages:
1) This patent can only be applied to databases as it needs to begin a new transaction to write the log messages for the first transaction. It can not be used to application servers.

If the second transaction fails no log data will be saved as the second transaction will rollback.

In a different way from the method above, we use an extra server to process log messages produced by business application servers. The main points of this invention include the following:

1) An extra lightweight server is setup on each physical machine within a cell for log writing.
2) Some exception criteria's and related trace/debug specificat...