[TA]z 2004-05-04 09:05:01 | Stephen, The memory occupied by the client tends to increase gradually over time. In some instances, MUON1 is occupying more than 330 MB of memory. *EDIT* I have just found clients where 736+ MB memory is taken by MUON1. |
Stephen Brooks 2004-05-06 02:19:46 | Hmm. I went around Muon1 with a memory-leak detector program a while back, but maybe I've messed something up in the interim. I needed to check out the RAM consumption anyway for v4.4 as there could be many more particles than before. [edit] What do you use for measuring process memory usage? I've found the readout in Task Manager does weird things, as if parts of the memory are being swapped from one counter into another. |
[TA]z 2004-05-06 06:38:49 | I use pslist.exe (which I believe uses the built-in windows performance monitor) to poll every machine on the network for the muon1 process and it returns info for the process.
|
Stephen Brooks 2004-05-06 09:38:11 | I've just run the leak test again and it appears I've got several small leaks, which I'm going to fix. Probably if they've been running as long as your machines appear to be running, that'll be enough time for those leaks to really add up. |
[TA]z 2004-05-06 10:55:27 | Aha! Admit it, you're secretly taking pleasure in the fact that you've utilized over 7 gigs (I've seen it as high as 10 gigs!) of my memory . Besides, I've known all along that you're secretly breeding mallard populations and tracking them with the muon1 program... It seems the program first determines the probability of two mallards engaging in the function "SEXT" based on their relative "SKSE" levels. If "SEXT" happens, you measure the "width" of the result to classify the new mallard as "big" or "small" M-U-O-N M allard U psurge O bservation N etwork |
[DPC]Stephan202 2004-05-06 15:47:51 | Things start to taste a little marmite flavoured over here. |
Stephen Brooks 2004-05-07 02:01:11 | You know, I was expecting boring replies this morning, for some reason. Instead it appears I have leaky mallards |
[OCAU] badger 2004-05-09 19:15:50 | quote: Oh that'll leave a mess. I have noticed that when I leave this machine running muon over the weekend, when I get back nothing will run (insufficient memory) or it has hung. This would explain why... I guess I need to restart the process every so often... |
[TA]Confused 2004-05-10 03:03:41 | I'm having this problem with some systems as well, running Windows 2000, which are on 24/7. I don't want to have to remove them from the project, but I feel that I have to, as the systems are slowing to almost a halt, and therefore running DPAD is not as transparent to the end users as I would like. Garry |
Stephen Brooks 2004-05-10 03:06:20 | That's OK for now - the other possibility is if you can figure out some way to get (e.g.) Task Scheduler to shut down and restart the Muon1 program every 12/24 hours or something. I'm trying to trace the memory leak(s) at the moment - I won't release v4.4 with known leaks. [update] Fixed 2 small leaks so far. There appear to be about 3 or 4 other small ones, plus one big and rather complicated one. |
[TA]z 2004-05-10 07:28:52 | good to hear thanks for the update |
Stephen Brooks 2004-05-10 08:19:26 | The "3 or 4 other small ones" turned out to all be caused by the same thing, which I've fixed. Now it's just the peculiar one to do with the decays. Trying to do memory management when the things you are managing randomly split and are also being dealt with in multiple threads has the potential to be a little tricky... BTW, the stats were down for a while today, also due to some code rearrangment (I foolishly recompiled genstats with buggy code, meaning I was obliged to fix those particular bugs and get it back online quickly). It seems to be fine now. [update] I've just fixed the big leak as well. All of these were correctable by essentially one-line modifications like inserting a "free" command somewhere. The trick was figuring out _where_... |
[TA]z 2004-05-10 09:44:55 | will you replace the v4.34 client with one that addresses the leakiness or wait until v4.4 to update the latest client available to us? |