Background Process Questions


#1

I am attempting to run a background Java/JRE listener process. After 15-30 seconds it gets killed.

Does DH have some kind of watchdog daemon that kills bg processes?

The start/stop/status script nohup’s and &'s the process. Environment variables have been exported in the parent. The main bg process has 11 java threads to it. Is there a limit on the number of background processes you may have?

I thought you might have to run bg process from your main server account, but in 24 hours the main server ftp hasn’t been converted to shell yet.

By default there are 8 points of nice added so that shouldn’t be the problem. I ran the script foreground/debug and watched the processes with TOP, and they aren’t taking any resources at all.

This is, alas, a chat system with a Java listener on the server and a client applet. Sales told me that as long as it wasn’t IRC I could give it a shot as long as system resources weren’t hogged. Could there be restrictions on the ports that can be used? Can ports be used at all?

There are two CRON jobs associated with the system as well. One that restarts the main processes at 4:00 a.m. (JAVA memory-leak precaution) and a watchdog that checks every 15 minutes to make sure it’s still up, and restart if not. Could these be the problem?

Support must understandably be busy with the datacenter move tonight as I haven’t received a response from the trouble ticket either.

Sorry to ask so many questions in one email, but the software is $500 and I want to make sure it works right before I plunk down more than double what the site cost. Any feedback from you old-timers out there would be very much appreciated.

Thanks…


#2

Just got a reply from support that verified they have a killer daemon running abolishing any expressions of originality.

Should I use my 30 day money-back and get out while I can?

I am furious at this point. Is there anyone left in the world that knows ANYTHING about customer satisfaction?


#3

[quote]Should I use my 30 day money-back and get out
while I can?

[/quote]

I can’t imagine why you ordered web hosting, if you wanted something else. The guarantee? A gift. Take it.

[quote]I am furious at this point. Is there anyone left in
the world that knows ANYTHING about customer
satisfaction?

[/quote]

Apparently not. MY idea of customer satisfaction is that a restaurant should give a good hamburger to the guy who orders a hamburger and a good 24-ounce porterhouse dinner to the guy who orders a 24-ounce porterhouse dinner. Apparently you think you should get a 24-ounce porterhouse dinner for the price of a hamburger.

It looks like Dreamhost does offer what you want. At http://www.dreamservers.com, they offer dedicated servers. But not at hamburger prices.


#4

Thanks, I needed that!

Of course you are correct deke. I should remember how happy I am just to have a shell account.

Now that I think about it, this burger did come with a lot of ‘fixins’!


#5

without getting too specific (for obvious reasons), i would mention that the number of processes is part of what the process killer evaluates, so if there’s a way to make the job fork fewer threads, that might help.

writing support (nicely) and asking about it might very well result in some sort of compromise being reached.

our aim is to prevent one user from jacking up the load on a machine, and to leave resources free for its primary purpose (ie serving our customers’ webpages). we’ve had a lot of problems with runaway cgi scripts that keep our pagers going off every 20 minutes. :>

i haven’t seen it cause problems for very many users; it appears that your case may be an exception.

we are working on making it a bit smarter as far as which processes are actually taking up too much cpu / memory and which ones aren’t.


#6

Well I found out what the problem is. Java 1.3 on Linux shows each Java thread as a separate process on the system. So my multi-thread background process looks like 12 are running. Each thread shares the same memory space but just looking at ps one would think it’s using 120mb of memory! Sun doesn’t seem to be responding to the many cries for help.

Anyone know if there is an older version of JRE on the system, maybe 1.1?

Thanks…


#7

yeah - linux ps is kind of stupid this way. bind 9 running threaded shows up the same way.


#8

Will, do you know if Java 1.1.8 is still out there anywhere? I cant find anything but 1.3.0 …

I’m hoping it’s still available and that will solve the problem. If it does not, then it is indeed the linux ps command and the way the the killer daemon interfaces with the OS.

The RealChat software that I want to run was carefully selected, and shouldn’t cause any resource problems. I’m working with the developer but we’re at a stand-still at this point.


#9

out there as in available or out there as in on our machines? i don’t know too much about java personally; java.sun.com still has 1.1.8 IIRC.


#10

Thanks Will!

Actually, thanks to deke setting my train of thought back on the right track and you Will for suggesting nice email, Nate sent me several great emails and I think we’ve gotten everything running now! :slight_smile: (thanks guys)

I tried JRE 1.4 which showed more threads than 1.3x! So I loaded 1.1.8 from blackdown and the process now only shows as one.

However, I’m still worried about memory leaks. I would rather run under 1.3 + but for now 1.1.8 is doing ok.

Thanks!