JForums Caches not working properly

[Migrated From Jforum.net]

Running JForum 2.1.8 on Oracle App Server 10g.x.x against Oracle DB 10g.x.x.

1 Some posts are not showing up until I clear the topics cache from the admin screen.
2 Forums do not show next-page links when there are more than page's worth of topics, until I add a topic and delete it.

Right now I'm doing these manual cache fixes all day long to keep users happy. I however am not happy.

I can disable the Topics Cache and get rid of problem 1, I think. It looks like I would have to recode PostAction.java to force rebuild of ForumRepository's cache for problem 2. Both of these aren't great solutions, but I'm slightly desperate.

Am I the only person having this problem?
[originally posted on jforum.net by bazoople]

[Migrated From Jforum.net]

Update - the recent topics list also has cache problems.

I am adding a fake topic every couple hours and deleting it to fix Recent Topics and make paging work properly.
[originally posted on jforum.net by bazoople]

[Migrated From Jforum.net]

Another update: Fortunately disabling the Topics Cache seems to take care of everything. We've got about 600 users posting 30-60 items a day and aren't having performance problems, so I'm good.


[originally posted on jforum.net by bazoople]

[Migrated From Jforum.net]

Strange all of these cache problems. I know some bugs, but not that much.

Anyway, good to know you're running fine by now.

For JForum3 all this code is being rewritten.

Rafael
[originally posted on jforum.net by Rafael Steil]

[Migrated From Jforum.net]

Only thing I can think of is that you're running in a clustered environment of some sort. Since the cache is memory based, one cluster's JVM cache object would be updated with a new post with the other cluster's JVM would not.
[originally posted on jforum.net by monroe]

[Migrated From Jforum.net]

monroe wrote:Only thing I can think of is that you're running in a clustered environment of some sort. Since the cache is memory based, one cluster's JVM cache object would be updated with a new post with the other cluster's JVM would not.

Nope, we actually set up a single-server environment for apps like this (i.e. that want to write to the hard drive). There are clustered servers in front of that, but I think they just reverse proxy back to it. Images always work perfectly, whereas in a cluster each server would have only part of the images I think.

It may have something to do with Oracle's JVM and how it handles synchronized access or some other obscure threading issue(s). It's their 1.5 JVM. They get it from some other company...

But yes all is working well now.
[originally posted on jforum.net by bazoople]