I've been monitoring the disk usage in my StateDir for some time now and I've found rather interesting stuff about it.
First the basics: it's an nfs shared statedir, used by four web servers. I'm using a ramdisk for it, since using an actual disk for it proved discouraging.
Here are the stats: measuring from 30.1.2004 – 18:25:00 to 2.2.2004 – 20:32:00 number of disk writes:9728300 number of disk reads : 100310
I can't imagine that these performance metrics are measuring what you think they are. There is no way that there is a 97/1 write/read ratio.
I would consider that your mirrored situation might be buffering/caching reads locally, and only writing through, so you only see a small fraction of reads. I don't really know enough about the mirroring technologies you are using to suggest more, but certainly these numbers are pretty suspicious.
When I go to get an example of stats from within MLDBM::Sync ( don't try this a
Any idea why things are the way they are and what could/should be done ?
If you want to get to how an Apache::ASP process does its read/writes to the MLDBM::Sync databases, I would fire up the apache web server under httpd -X mode ( for apache 1.3.2x ), while running under strace capturing the system calls that you are interested in. This assumes you are using linux, but I am sure there are other system call trace utilities for your OS. This would help isolate the stats from the web server side, vs. stats that are particular to your clustering/mirroring software.
Regards,
Josh
________________________________________________________________ Josh Chamas, Founder phone:925-552-0128 Chamas Enterprises Inc. http://www.chamas.com NodeWorks Link Checker http://www.nodeworks.com
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
