Description:
If mysqlbinlog finds a corrupted binlog event (e.g. binlog event checksum failure), it aborts if the binlog file was properly closed. However, it silently exits with exit code 0 if the corrupted event was found on the active (marked as LOG_EVENT_BINLOG_IN_USE_F) binlog files.
But in some cases, we may want to know if binlog files have corrupted events on in-use (not properly closed) binlog files as well, rather than silently ignoring errors.
Why not adding a command line argument to control whether to ignore event errors on in-use binlog files?
How to repeat:
Generate binary logs with binlog checksum enabled. Edit the newest (opened) binary log and edit some events (e.g. changing table name in table_map log event). Run mysqlbinlog --verify-binlog-checksum to see if errors are printed and check exit code.
5.6: mysqlbinlog prints "Event crc check failed! Most likely there is event corruption." on event checksum error, but it exits with 0.
Log_event* ev = Log_event::read_log_event(file, glob_description_event,
opt_verify_binlog_checksum);
if (!ev)
{
/*
if binlog wasn't closed properly ("in use" flag is set) don't complain
about a corruption, but treat it as EOF and move to the next binlog.
*/
if (glob_description_event->flags & LOG_EVENT_BINLOG_IN_USE_F)
file->error= 0;
8.0.17 mysqlbinlog does not print anything on event checksum error, and exits with 0.
Log_event *ev = mysqlbinlog_file_reader.read_event_object();
if (ev == NULL) {
/*
if binlog wasn't closed properly ("in use" flag is set) don't complain
about a corruption, but treat it as EOF and move to the next binlog.
*/
if ((mysqlbinlog_file_reader.format_description_event()->header()->flags &
LOG_EVENT_BINLOG_IN_USE_F) ||
mysqlbinlog_file_reader.get_error_type() ==
Binlog_read_error::READ_EOF)
goto end;
Suggested fix:
Adding a new command line argument in mysqlbinlog to control whether to ignore event errors on in-use binlog files.
Description: If mysqlbinlog finds a corrupted binlog event (e.g. binlog event checksum failure), it aborts if the binlog file was properly closed. However, it silently exits with exit code 0 if the corrupted event was found on the active (marked as LOG_EVENT_BINLOG_IN_USE_F) binlog files. But in some cases, we may want to know if binlog files have corrupted events on in-use (not properly closed) binlog files as well, rather than silently ignoring errors. Why not adding a command line argument to control whether to ignore event errors on in-use binlog files? How to repeat: Generate binary logs with binlog checksum enabled. Edit the newest (opened) binary log and edit some events (e.g. changing table name in table_map log event). Run mysqlbinlog --verify-binlog-checksum to see if errors are printed and check exit code. 5.6: mysqlbinlog prints "Event crc check failed! Most likely there is event corruption." on event checksum error, but it exits with 0. Log_event* ev = Log_event::read_log_event(file, glob_description_event, opt_verify_binlog_checksum); if (!ev) { /* if binlog wasn't closed properly ("in use" flag is set) don't complain about a corruption, but treat it as EOF and move to the next binlog. */ if (glob_description_event->flags & LOG_EVENT_BINLOG_IN_USE_F) file->error= 0; 8.0.17 mysqlbinlog does not print anything on event checksum error, and exits with 0. Log_event *ev = mysqlbinlog_file_reader.read_event_object(); if (ev == NULL) { /* if binlog wasn't closed properly ("in use" flag is set) don't complain about a corruption, but treat it as EOF and move to the next binlog. */ if ((mysqlbinlog_file_reader.format_description_event()->header()->flags & LOG_EVENT_BINLOG_IN_USE_F) || mysqlbinlog_file_reader.get_error_type() == Binlog_read_error::READ_EOF) goto end; Suggested fix: Adding a new command line argument in mysqlbinlog to control whether to ignore event errors on in-use binlog files.