I’m inspecting a rather large server cluster. There are 15 odd servers running instances of an ERP. The various instances tend to have specific uses & then there are also 2 high traffic web sites hosted with apache & mod_jk for static content.
All dozen server use a single database. Which I’m looking into the appropriate High Availability, Load Balancing,Replication solution for.
I’ve been keeping an eye on the Sys Admin guy for a couple of days and all he seems to do is restart the server when an end user complains of slowness or unresponsiveness.
In fact some end users send emails saying, pls restart the server.
I’m a software engineer but my clients have asked me to leave the application layer for now at least the application itself & try to boost performance via the layers beneath, container/appserver, io, network, hardware I suppose.
Anyway can anyone give me an indication of why slow down occures at high usage & why a restart would help, there must be a way to achieve the result of a restart before a restart is needed or an end user had reason to complain.
I realize this may be a dumb question, I feel as if I should know. It might have to do with IO and sockets along with the threads working with them. But I need more insight. As I’ve gained development experience I’ve got closer to ground zero, I suppose the lowest level of the application, after which I suppose comes the network stack then hardware. I’ve never had a mental image of what happens beneath where the web app context is created and added to tomcat embeddable.
A bit of background, Tomcat 5 embedded, along with Apache Geronimo. I know for a fact the this version of the app is old, hasn’t upgraded when the project that birthed it was still around. It has cache issues & thread handling issues, which were both improved in later additions but upgrading right now isn’t an option.
The cause in Tomcat is almost always running out of memory.