A

If you are running a Linux laptop that uses NetworkManager, you can get a list of every wireless network that NetworkManager knows about by doing

gconftool-2 --all-dirs /system/networking/wireless/networks

If you want to delete one, do

gconftool-2 --recursive-unset /system/networking/wireless/networks/BADNETWORKNAME

I find this useful when I'm in a venue, such as a tech conference, where there is both a free conference wireless network, and a paid wireless network run by the convention center. If you ever accidentally associate with the paid network, then whenever the signal strength of the paid network is higher, NetworkManager will prefer it to the free one. The fix is to delete the paid network data from the gconf database.

So, I had written a while back: “We currently have a DataMemory of 4GB and IndexMemory of 2GB. Based on the crude methods we have to monitor it, I think we are at about 40% capacity.” Boy, I was wrong.

After that post, I started looking at this more in detail because we were considering buying more RAM “just in case”. I figured out how to use the super secret command “all dump 1000″. The command is not documented in the MySQL documentation that I could find. I did find it in the NDB API documentation before writing this post however. Not sure why I could not find it before.

For those that still don’t know how to use it, simply type “all dump 1000″ from your management console. Then check your cluster log files …

A couple of weeks ago, a friend asked about replication on MySQL 4.1.7. I’ve worked with replication in the past, just a quick and dirty job on MySQL 5, and soon forgot about it. This time, I wanted to do it on MySQL 4, and make sure I take good notes for my own benefit. If it can help somebody else, all the better. The official documentation is here. It took a little time to wade through it.

The process below is used to replicate all databases on the master to the slave(s). I will talk about replicating only certain selected databases in a future post.

1. At the master server, check if binary logging is on. Use show variables like ‘%bin%’ to check if binary logging is on or not. If it is not on, turn it on by adding log-bin under [mysqld] section in …

One of these years I will finally get a talk accepted for OScon and actually be able to attend. I always hear good things about the conference and it seems like my company almost has its own track this year.

For those of you who are attending, be sure to attend some of the great talks being delivered by my friends and colleagues:

Improving Performance by Profiling PHP Applications
Next-Gen Interaction with Ajax, Comet, and PHP
Advanced Production Troubleshooting
Scalable Internet Architectures …

In a talk with Raph Levien today he said:

[basically] “… the metric for the success of an open source software project is whether the author now works at Google.”

… nearly everything I’ve ever written fails to pass this test.

Hopefully someone from Google’s M&A department can help me remedy this situation. (hint hint)

Over the weekend we migrated one of our tables from MyISAM to InnoDB. Generally speaking we’re pretty happy with most of our InnoDB migrations.

InnoDB generally uses 2-3x more data on disk but the write ahead log (WAL) buys us so much more added performance that it’s generally worth the switch.

This migration however was a bit more problematic. Importing the data from mysqldump as MyISAM into the new box only took about 2 hours. Converting the data to InnoDB has so far taken 24 hours and using 21G vs the original 6.1G for MyISAM.

I should also note that it’s not complete yet and it’s 3.5x larger than our original MyISAM install.

Clearly the performance will be worse in our situation for InnoDB because I don’t think the buffer pool efficiency will make the write ahead logging fast enough to beat MyISAM on the same hardware.

We’re not the only people who have …

As I felt the itch to do some quick hacking yesterday, I decided to provide an RPM spec file for the MySQL proxy. The changes have been commited to the SVN trunk now and I added some hints to the INSTALL file on how to perform an RPM build.

Here is a quick summary of how to convert the current SVN code into an installable RPM. You build environment needs to fulfill a few additional prerequisites (a gcc compiler and the C library header files are taken for granted here), I added the versions I used on my openSUSE 10.2 system for reference:

• autoconf 2.56 or newer (autoconf-2.60)
• automake 1.9 or newer (automake-1.9.6)
…

For the curious, the slides for the eLiberatica 2007 presentations are now up at http://eliberatica.ro/2007/ - the slides include presentations from eZ CEO Aleksander Farstad, Brian Behlendorf, FSF Europe President Georg Greve, Ubuntist Kurt von Finck and MySQL co-founder Monty Widenius (and yours truly, but the slides are a bit weak - check out my …

Samsung has announced a new line of 64G SSDs for use in consumer laptops. Nice to see innovation but I REALLY want to see more SSDs in the server space.

I JUST want to use a CHEAP 2-5G SSD for write ahead and binary logs. I just need 1k write ops per second with a SATA interface. Then I could drop them in cheap commodity servers and get a significant performance boost. My disks would still use software RAID 0 (at least until I replace this partitioning) but only store data.

Give me one of these for $50 to $100 and I’ll buy one with ever server.

The fsync’s required for writing to disks on INNODB can really hurt performance. My bet …