I have been running a public MySQL server for over 4 years. The system was a 1GHz box with 256MB of RAM. MySQL puts each database into a seperate directory in a single "data" directory. Once the old system reached about 10K databases the connection times increased 10X or more. I attributed the speed problems to a possible filesystem limitation with a large number of files. That system ran RH9 and either had ext2 or ext3 with no htree patch. Recently I bought a new 2.4GHz server with 1GB RAM and much faster drives. I installed FC4 and am running a 2.6.14 kernel. I thought that even if my problems were not completely solved, I would at least not see connection times increase until I had many more directory entries (wild guess - 20K). Since I am writing, you can guess that is not the case. At 13K directory entries, I am still seeing significantly slower connection times even with the faster hardware and newer software. I'm limited in what I can move because MySQL expects everyting in the "data" directory. My questions are: 1) I don't know much about htree, but I recall that it is supposed to help in this situation. How do I tell if it is in use on my system. Is it a kernel module or is it compiled into ext3? 2) Are there other options for tuneing ext3 performance? Thanks everyone, Gary Huntress
On Apr 01, 2006 09:53 -0500, Huntress Gary B NPRI wrote:> Since I am writing, you can guess that is not the case. At 13K > directory entries, I am still seeing significantly slower connection > times even with the faster hardware and newer software.You need to run "e2fsck -f -D" when the filesystem is unmounted, to build htree indices for your large directories.> My questions are: 1) I don't know much about htree, but I recall > that it is supposed to help in this situation. How do I tell if it is > in use on my system. Is it a kernel module or is it compiled into ext3?"dumpe2fs -h <device> | grep -i features", it's part of stock 2.6 ext3. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.