congratulations by your blog. Is interesting and helpful. I have a question about:

tmp_table_size

you said that “will grow as MEMORY table before it will be converted to on disk MyISAM table”. I read more of tmp_table_size, one of another definition was “Maximum size of in-memory tables”. Someone can give more ideas about what is a in-memory tables? are a temporary tables which are created as .MYI file on MySQL sometimes?

Thanks

Download EBooks Freely in .pdf,.chm formats and ready for interview questions….

http://hotsoftwareslist.blogspot.com/2007/08/download-ebook.html

Download Opensource CMS Freely

http://phpfunda.blogspot.com/

Project management System in PHP – MySQL – http://phpfunda.blogspot.com/2007/09/project-management-system-in-php-mysql.html

content management systems – http://phpfunda.blogspot.com/2007/08/content-management-systems.html

You do not need to execute mount none /mnt/hugepages -t hugetlbfs
for mysql

all what you need is
echo 2000 > /proc/sys/vm/nr_hugepages

or equal command for your Linux (what do you use btw ? )

As I see your command was not executed successfully:

HugePages_Total: 0
HugePages_Free: 0
Hugepagesize: 2048 kB

There should be
HugePages_Total: 2000

There can be several reasons:
1. Your memory is fragmented, so kernel can not allocated non-fragmented part with requested size,
can you try allocate just after reboot ?
2. echo 2000 > /proc/sys/vm/nr_hugepages does not work on your box

See also Configuring Huge Pages in RHEL 3/4 here:
http://www.puschitz.com/TuningLinuxForOracle.shtml#ConfiguringHugePagesInRHEL4

I configured already a kernel to enable huge pages. I can use huge pages with a size of 2MB.

I made a script that does the following:
mount none /mnt/hugepages -t hugetlbfs

I did that because I saw with the monitoring tool strace that mysql uses mmap system calls.
Here you can see the output of
strace -c -f -o strace.txt -p 4721 (4721 is the pid of mysql)

% time seconds usecs/call calls errors syscall

84.78 18.000.427.471 10362 1737133 144871 futex
5.11 1.084.404.926 22078 49117 1 select
4.89 1.037.780.473 393 2637973 274671 read
1.25 264.358.705 46855 5642 rt_sigtimedwait
1.07 226.303.746 96 2368066 time
0.89 189.576.561 294 645392 sched_yield
0.42 90.140.144 80 1126118 rt_sigprocmask
0.42 89.731.199 88 1013972 1013972 sched_setscheduler
0.30 63.858.894 112 572717 4012 write
0.20 42.552.311 185 230154 munmap
0.20 42.525.781 75 564870 fcntl
0.16 32.911.366 135 243164 mmap
0.11 23.633.754 29 807912 lseek
0.08 16.839.725 62 271100 open
0.05 10.935.959 21 530205 pread
0.02 5.296.405 20 268102 265100 access
0.02 4.758.093 17 272612 close
0.01 2.227.735 154 14467 mprotect
0.00 0.484223 124 3906 pwrite
0.00 0.445093 404 1102 fsync
0.00 0.430290 72 6000 1500 lstat
0.00 0.372081 124 3002 unlink
0.00 0.347109 61 5701 tgkill
0.00 0.334789 74 4503 setsockopt
0.00 0.306899 54 5639 alarm
0.00 0.269266 60 4500 4500 readlink
0.00 0.132078 88 1501 getpeername
0.00 0.113291 75 1501 accept
0.00 0.112548 75 1504 shutdown
0.00 0.095832 64 1501 getsockname
0.00 0.004834 44 109 clone
0.00 0.001256 60 21 brk
0.00 0.000094 47 2 gettimeofday
0.00 0.000046 23 2 rt_sigaction
0.00 0.000028 14 2 2 rt_sigreturn
0.00 0.000009 9 1 clock_gettime

100.00 21.231.713.014 13399213 1708629 total

I also put the parameter larges-pages in the my.cnf and when I check in mysql I saw the following:

show variables like ‘%large%’;
+———————+———+
| Variable_name | Value |
+———————+———+
| large_files_support | ON |
| large_page_size | 2097152 |
| large_pages | ON |
+———————+———+
3 rows in set (0.00 sec)

Now I have a few questions:

- Did I do something wrong here? Or am I forgetting something?
- Can I assume that the large_page_size here is 2097152 kB? So it uses 1048 pages of 2M?
- Must I use the following or is mysql doing that when it starts:
Use the following command to dynamically allocate/deallocate hugepages:
echo 2000 > /proc/sys/vm/nr_hugepages
- and the most important question for me is: How can i see that mysql use this huge pages.

Information about the server that I use:

Linux DL585 2.6.15.1-huge-page #1 SMP Thu Nov 23 22:37:15 CET 2006 x86_64 GNU/Linux

cat /proc/meminfo
MemTotal: 8012512 kB
MemFree: 6006088 kB
Buffers: 7484 kB
Cached: 1019872 kB
SwapCached: 0 kB
Active: 962876 kB
Inactive: 1003952 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 8012512 kB
LowFree: 6006088 kB
SwapTotal: 975112 kB
SwapFree: 975112 kB
Dirty: 0 kB
Writeback: 0 kB
Mapped: 948948 kB
Slab: 18344 kB
CommitLimit: 4981368 kB
Committed_AS: 1555036 kB
PageTables: 3248 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 3956 kB
VmallocChunk: 34359734159 kB
HugePages_Total: 0
HugePages_Free: 0
Hugepagesize: 2048 kB

I know I ask a lot, but it should be a great help for me if anyone could help me out.

Thanks,Anja

You need to start mysql with –large-pages options
http://dev.mysql.com/doc/refman/5.0/en/server-options.html

But before, you have to configure your Linux to enable huge pages.
for example you can read hugetlbpage.txt :
http://www.mjmwired.net/kernel/Documentation/vm/hugetlbpage.txt

However take into account RedHat uses its own way to enable huge pages, perhaps others vendors too.

I’m trying to test mysql with large-pages. I compiled a kernel so it support huge pages. Now I want to use large-pages in mysql. How do I use the large-pages? How do I know that mysql use this huge pages? Is there a way to see that?

Anja (Student)