Zach Dennis
3/23/2006 11:36:00 PM
Berlin Brown wrote:
> That is interesting. I like that approach. How does the ruby client
> code work.
The client just opens a Socket connection to the server, and sends it instructions. The server takes the instructions, and the
server process creates a new thread and does whatever it needs to do based on instructions. We make the following assumption:
* the server knows how to parse basic instructions so it can load the correct library/worker and pass on worker-specific
instructions to it
* the client understands the format of instructions
> Basically, does it timeout or cause a lot of issues.
A handful of times it was reported that it didn't correctly generate the report for the end user, however, it was hit half a
million times in about 2 weeks and it's hard to prove a handful of times something not working. Although I am investigating. =)
The biggest issue we've seen is memory consumption. We use this tactic for generating CSV reports from a database. Some CSV
reports which include 45,000 or greater (upward of 275,000) records can really suck up the memory. We have avoided this as being a
huge problem by doing two things:
1. beefing up server memory
2. restarting our worker processes (and daemon processes) periodically if memory consumption gets way to high.
There are some changes I am making to my system to fix these initial issues, if interested contact me off list...
Zach