WAS ist das? RANT

All the app servers and servlet containers I’ve seen these days the IBM’s Websphere Application Server tops them all. The most unexpected, complicated and difficult to understand by far.

Lets start by trying to find out what the server is doing at a given point. The startServer.sh will echo the name of the log file that you can tail. Great! At one point you understand that there’s a lot more going on in the background.

You do a quick find -mmin 3 and find out that do get an overview you have to tail about 5 logs: startServer.log, stopServer.log, SystemErr.log, SystemOut.log, server1_exception.log. Divide and conquer but how am I supposed to keep an eye on them? How many pages of documentation do I have to read to monitor the correct one?

I won’t even start on how to find the administrative console or what services are started on which ports. Once you find it and start deploying an application you understand that exploded format is a big no-no, no option for that. Ok, let’s copy or link the application to the installedApps folder. Nope, does not cut it. Nothing happens, none of the 5 logs show anything.

Just zip -r ../app.ear * and deploy the ear file. It will be unpacked to installedApps and off you go. Well actually you still have to zip even the WAR files inside the EAR for the deployment to be a success.

Now you configure your IDE to generate webapp’s classes to the deployed application. You see that the application starts reloading after every change. Probably somekind of auto loading. You go to the administrative console, find “Class loading and update detection”. This must be it. The checkbox is _un_ticked??? Why is it reloading on every update then? I don’t know, but just enable the damn thing and use a really large polling interval and it will all go away.

I can actually go on but I’m probably just wasting time. By the way, Glassfish’s profile/domain is 6.7 MB, my IBM’s profile/domain is 458MB. Go figure.

Tags:

This entry was posted on Friday, February 22nd, 2008 at 8:31 am by Toomas Römer and is filed under opinion. You can follow any comments to this entry through the RSS 2.0 feed. You can leave a comment, or trackback from your own site.

  • Vladimir Shor

    I have yet never seen a good and usable piece of software produced by IBM.

    ClearCase – monstrous versioning system with totally unusable user interface. Seems quite unstable as well, as I’ve seen a department of four people just maintaining it for a not so big development group (I guess not more that 100 people) and the number of active projects there is about 20. Ok, couple of them are quite a big bastards, but that’s no excuse for the money IBM is asking for it. Has a rich client (uses at least 3 ports to talk to the server), remote client – built on top of Eclipse (runs over http) and a Eclipse plugin (you should see comments about it in eclipseplugins community:)). None of these software pieces is usable.

    ClearQuest – bugzilla connected with ClearCase. Even for this simple task its rich client uses at least 2 ports to talk to the server. Web-client is very flexible and thus very complicated to understand and configure. The same big ugly unusable monster.

    IBM software for ThinkPad notebooks – please! Why default windows components are not good enough and why invent some ‘Productivity Center’ to make simple things complicated?

    I guess the one rare OK thing from IBM is Eclipse. But it is not in IBM now, so thats why, I guess, it is more or less usable nowadays.