Comment 2 for bug 1007446

Revision history for this message
Miguel Angel Nieto (miguelangelnieto) wrote :

This is the process if there is a xtrabackup_suspended on tmpdir.

1- innobackupex starts
2- innobackupex launches xtrabackup
3- innobackupex sees that there is a xtrabackup_suspended, so it thinks that xtrabackup finishes copying *.ibd files. Therefore, both, xtrabackup and innobackupex start their tasks at the same time

[01] Streaming /m01/mysql/ibdata/ibdata1

120924 15:37:57 innobackupex: Continuing after ibbackup has suspended
120924 15:37:57 innobackupex: Starting mysql with options: --host='127.0.0.1' --unbuffered --
120924 15:37:57 innobackupex: Connected to database with mysql child process (pid=5652)
>> log scanned up to (4110378914)
120924 15:38:02 innobackupex: Starting to lock all tables...
>> log scanned up to (4110378914)
>> log scanned up to (4110378914)
[01] ...done
[01] Streaming ./employees5/dept_emp.ibd

As we can see here, innobackupex starts to lock tables but xtrabackup didn't finish.

4- innobackupex finishes copying *.frm files and delete xtrabackup_suspended

120924 15:38:39 innobackupex: Finished backing up .frm, .MRG, .MYD, .MYI, .TRG, .TRN, .ARM, .ARZ, .CSV, .CSM and .opt files
innobackupex: Resuming ibbackup
[01] ...done
[01] Streaming ./employees10/dept_manager.ibd

5- xtrabackup finishes copying *.ibd files and creates a xtrabackup_suspended

[01] Streaming ./employees10/departments.ibd
[01] ...done
>> log scanned up to (4110378914)

6- xtrabackups waits to innobackupex to delete xtrabackup_suspended, but innobackupex finished all its task before, so nobody is going to delete that file. therefore, xtrabackup stays in "log scanned up to" forever.

>> log scanned up to (4110378914)
>> log scanned up to (4110378914)
>> log scanned up to (4110378914)
>> log scanned up to (4110378914)
>> log scanned up to (4110378914)
....

This problem could also cause a disk space problem, because it will stay copying redo log *forever*.