MySQL Bugs: #16787: DD: CREATE LG with INITIAL_SIZE 3G returns error after 40 min.

Bug #16787	DD: CREATE LG with INITIAL_SIZE 3G returns error after 40 min.
Submitted:	25 Jan 2006 17:00	Modified:	2 Mar 2006 17:59
Reporter:	Nikolay Grishakin	Email Updates:
Status:	No Feedback	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S3 (Non-critical)
Version:	5.1	OS:	Linux (Linux)
Assigned to:	Assigned Account	CPU Architecture:	Any

Description:
The following test case returns error after 40 min: "ERROR HY000: Failed to create UNDOFILE", no error code. Why should we wait for such long time if INITIAL_SIZE limit is 2G. If 2 GB is limit, then it should return an "unsupported" error right away.
 
 CREATE LOGFILE GROUP lg1
 ADD UNDOFILE './var/undofile.dat'
 INITIAL_SIZE 3G
 UNDO_BUFFER_SIZE = 17M
 ENGINE=NDB;

How to repeat:
see above

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/1667

Hi,

I tested this on my machine wo/ problem.
It took ~50s to create a 3G file.

1) These kind of limits are very dependant on OS, configure swithes libc etc.
2) If you running multiple nodes on 1 machine with 1 disk...it will cause disk trashing
    as two processes will try to create a huge file as fast as they can.
    This would explain why it took so long
3) Are you sure that you just didnt get disk full

---

I however found and fixed a bug when creating huge datafiles.

With 3G test fails with error 1506: Failed to create DATAFILE. You told me before you were able to create 3G but I’m using latest code from mysql-5.5-new and it fails.

You can use my test machine ndb15:/ngrishakin/mysql-5.1-new/mysql-test.
Test case to run:
[ndbdev@ndb15 mysql-test]$ ./mysql-test-run --do-test=ndb_dd_limits

ndb15> df -h
/dev/cciss/c0d0p3     9.2G  8.5G  250M  98% /
/dev/cciss/c0d0p1     190M   22M  158M  12% /boot
/dev/shm              3.9G     0  3.9G   0% /dev/shm
/dev/cciss/c0d1p1      67G  676M   63G   2% /space
/dev/cciss/c0d2p1      67G  181M   64G   1% /data0
/dev/cciss/c0d3p1      67G  181M   64G   1% /data1
/dev/cciss/c0d4p1      67G  181M   64G   1% /data2
ndbmaster:/ndb        130G   59G   65G  48% /ndb

If your running on "/" it's not big enough.

Further more, if you running "default" mysql-test-run.
It will start two nodes, both on same disk.
Creating 2 big(3g) files, will cause sever disk trashing.
If you want to test this, it's probably smarted to handle the configuration your self.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".