Ben Coddington told me:
We’ve had some loading lately when bots crawl /landscape. Looks like /landscape often opens its own files through the web server, which clogs things up.
The worst requests are like these:
/landscape/search/details.php?ls=55237&sequence=000&set_seq=3533&imageSet=1367046201-517b7839a64eb&AddRel=
They make several self-requests to load image files through the web server, even though only HTML is returned. I got to looking at the code a bit, and it looks like there’s a test to see if an image exists before writing out the HTML to load the image, but that test actually loads the image into the web server.
There are some other optimizations that can be done here as well; there are some very long-running database queries. Would you be willing to work with me to make the site run a bit better? Right now, indexing of this site causes all the SAA pagers to make noise – and that happens a couple times a day.
I made some code changes but was reminded that it was doing what it was doing because of php “Safe Mode” and we needed to change fuel permissions and ownership on tens of thousands of file for the code changes to work. This prompted Ben and Mike Austin to conclude:
Might be nice to get this thing on a dedicated server — we could turn off safe_mode, and the filesystem access would be much faster. Then it could age gracefully without UVM dancing around trying to fit it into our architecture over and over again.
How would you feel about that?
Turns out it didn’t matter how I felt, it is happening anyway. And of course, it isn’t all smooth sailing — walking the website now, finding and squashing bugs.
as far as sql goes,The most common query — the quick search — was performing a full text query on a set of fields which did not match the fields in the fulltext index. rebuilt the index, and queries formerly taking 3-4 seconds are now clocking in at 2.1ms .
