| # | Change | User | Description | Committed | |
|---|---|---|---|---|---|
| #4 | 9732 | Lester Cheung | Renamed direcotry "track2sql" to "log_analyzer". | ||
| #3 | 8061 | Lester Cheung |
Swapping NULL characters with slashes instead of spaces as it turns out that the problem was casused by a bug in P4V sending version strings containing NULL chars. |
||
| #2 | 8059 | Lester Cheung |
Replacing NULL characters with spaces. Some files log files contains NULL characters and track2sql would happily convert them into SQL statements. This may cause problem when we insert the generated statements into the database. For example: $ xzcat log.gz| php track2sql.php - -| mysql mydb ERROR 1064 (42000) at line 73164: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near '07d4e0a7-1b46-d94f-662d-a69a1ee2266f','trigger',54,0,53,1,0,0,1,1448,0,0,0,1,0,0' at line 1. ... which is not very helpful as it's not referring to the SQL statement which contains the NULL chars. |
||
| #1 | 8058 | Lester Cheung | Branching Steward's track2sql locally | ||
| //guest/stewart_lord/track2sql/track2sql.php | |||||
| #16 | 7621 | Stewart Lord |
Updated create table statements to use the IF NOT EXISTS condition. This allows SQL to be fed into an existing database without errors/warnings. |
||
| #15 | 7620 | Stewart Lord |
Minor update to Track2SQL to avoid timezone unset issue and split() function deprecated in PHP 5.3. |
||
| #14 | 7338 | Stewart Lord |
Rolled-back change 7209. This removes the endTime column from the process table. Track2SQL no longer attempts to analyze 'completed' entries. This change was influenced by three factors: - Analyzing 'completed' entries significantly degrades performance (about 2.75x slower in my tests). - In some versions of PHP (5.2.8) the strtotime() function suffers from a memory leak. - The 'completed' entries are not part of Vtrack output. |
||
| #13 | 7209 | Stewart Lord |
Integrating an enhancement from Michael Shield's guest branch. Track2SQL now records the end time of each process (when it is reported). This information is reported for every completed process when -vserver=2|3 logging is enabled. If verbose server logging is enabled this is more reliable than start 'time' + 'lapse' because (by default) lapse is only reported when it exceeds a certain threshold. If, however, vtrack=1 is set then lapse time will be reported for every command. Note: this change brings a schema change. It adds a 'endTime' column to the process table. |
||
| #12 | 7198 | Stewart Lord |
It is now possible to accumulate output from multiple invocations of track2sql in a single database. Previously, the processKey was a incrementing value that always started at zero. Therefore, processKeys from separate runs of track2sql would collide if inserted into the same databse. Now, track2sql uses a 36 character universally unique identifier (UUID) for each process. UUIDs give us reasonable confidence that the process keys will never collide. One consequence of this change is that the schema is slightly different. The type of the processKey column is now a 36 character varchar instead of a integer. Another side-effect of this change is that the output of track2sql cannot be predicted and will always be different even when processing the same log file. Track2sql performance is largely unaffected, however, insert and select performance may degrade somewhat due to the larger key size. |
||
| #11 | 7197 | Stewart Lord |
Fixed a bug where incomplete process lines such as: 'journal rotation', produced invalid SQL output. Log entries with process lines that fail to pass a basic sanity check are now ignored. The added check is rather crude and therefore performance is largely unaffected. |
||
| #10 | 7193 | Stewart Lord |
Fixed a bug where 'pages reordered' data could be mistaken for 'pages i/o' data, if the 'pages reordered' line was not preceded by a 'pages i/o' line. |
||
| #9 | 7104 | Stewart Lord |
Follow-on to 7098. The quote() function now wraps and escapes the given string. |
||
| #8 | 7098 | Stewart Lord |
Fixed a sqlite incompatibility. Backslash is not a valid way to escape quotes in sqlite. Quote-quote ('') is valid in both mysql and sqlite. Replaced use of addslashes with a quote() function that replaces occurrences of single-quotes with two single quotes. |
||
| #7 | 6424 | Stewart Lord |
Updated track2sql disclaimer. Addded a link to the readme file from the script itself. |
||
| #6 | 6379 | Stewart Lord |
Fixed a bug where track2sql failed to properly extract some table usage data if 'pages reordered' were reported. |
||
| #5 | 6289 | Stewart Lord |
Minor update to track2sql. - Added version and usage information. Can be viewed with -v, -V or -h. - Added error handling for the case of a non-existent input file or a empty input file. - Removed 'drop table if exists' statements from the table creation SQL. |
||
| #4 | 6010 | Stewart Lord |
Fixed a bug where track2sql failed to properly parse lock times in 2007.2 log files. This was due to a small change in the log file format. |
||
| #3 | 5889 | Stewart Lord |
Modified create table statements to use signed columns instead of unsigned columns. This avoids subtraction problems that can occur in some versions of MySQL when SQL_MODE is not set to NO_UNSIGNED_SUBTRACTION. Main() now sets error_reporting to E_ALL & ~E_NOTICE to suppress notices. |
||
| #2 | 5858 | Stewart Lord | Added disclaimer to script. | ||
| #1 | 5857 | Stewart Lord | Initial add of track2sql to the public depot. | ||