[Opensim-users] Odd but persistent symptom [text][bayes]
John Hopkin
opensim at jfhopkin.karoo.co.uk
Mon Jan 26 14:47:28 UTC 2009
Dirk Krause wrote:
>it doesn't really sound like this bug here
> http://opensimulator.org/mantis/view.php?id=2946
>but maybe you want to make sure that you set the number of threads high as proposed there
> export MONO_THREADS_PER_CPU=150
>with '150' being maybe a number that is too high.
Thanks - trying that now, and I'll report back. If there are any
other Mono tweaks that may improve things, I'd be interested; through
lack of better knowledge I simply use 1.9.1 out-of-the-box.
>What esp does not sound like this bug, is the console freeze. This is something that I didnt experience IIRC.
>
>Having said that, with freezing applications I always suspect hardware issues, namely faulty RAM.
True, and what I first suspected too, but moving onto another machine
entirely didn't make any difference in this case.
>-----Ursprüngliche Nachricht-----
>Von: opensim-users-bounces at lists.berlios.de im Auftrag von John Hopkin
>Gesendet: Mo 26.01.2009 01:43
>An: opensim-users at lists.berlios.de
>Betreff: [Opensim-users] Odd but persistent symptom [text][bayes]
>
>Pretty frequently, I've been experiencing lock-ups with OpenSim.exe
>that defy all attempts to diagnose by elimination. It's been like
>this for several weeks, across various releases up to and including
>8131. I'll briefly describe the symptoms here in the hope that
>someone might be able to point me in the right direction.
>
>1. When it happens, the server is usually up and running, and it's
>much more likely to happen when two avs are logged in (I've not tried
>any more than two) than with one. It can also happen with only one,
>or with none, and sometimes happens while scripts are still loading
>after starting the server, sometimes after an hour or so of running.
>Sometimes it can be left on overnight with no logins and it'll be
>fine.
>
>A typical scenario would be one av being logged in without problems
>for quite a long time; when a second av logs in, the server locks up
>within seconds. This, however, is not a completely universal rule,
>just the most common of the many scenarios.
>
>2. What happens is this: the region server stops responding at the
>command line, and nothing more will happen until it's killed off.
>Clients get logged off after timeout. On a system monitor, one
>virtual CPU (it's a P4, so single-core, but with hyper-threading there
>are two logical CPUs) is going flat-out at around 85-95%, the second
>CPU idling at around the normal 30%. This persists until the process
>is killed. Memory usage continues as normal. around 30-50% of the 2GB
>total.
>
>3. The server is running in grid mode, with UGAIM services provided
>by OSGrid, on a Ubuntu 8.10 server with Mono 1.9.1. There are four
>regions serviced by one copy of OpenSim.exe. Scripting is XEngine,
>physics ODE/Meshmerizer. There are around 2200 prims in all, with
>perhaps 400 mostly idle scripts.
>
>4. Eliminating each of the regions in turn seems to alleviate the
>problem to a certain degree - most of the time, any single region may
>be run with less problems than all four, but there is little pattern
>to this in the longer term. In the short term, it may appear that a
>particular region is at fault, but after a day or two the situation
>may change.
>
>5. Disabling scripting doesn't seem to have much, if any effect. No
>script is doing anything particularly exotic, and all are normally
>waiting for events.
>
>6. Another thing which sometimes (but not always) helps is to remove
>the mostly recently-rezzed prims.
>
>7. I've tried reinstalling Ubuntu, and I've also moved the whole
>thing from one machine to another during this time. OpenSim has been
>upgraded roughly weekly to the latest stable version hosted by OSGrid.
>
>It's hard to eliminate anything, because the situation changes so
>often; it may appear to run fine for several hours, with two avs
>making moderate use of the sims for building, etc, then it may barely
>run at all.
>
>If any of this looks familiar to anyone, I'd be very grateful for any
>help. I think you can imagine how frustrating this has been for
>myself and my partner, being almost completely unable to progress with
>our plans for building the regions into the land of our dreams. And
>it's particularly galling because of the lack of consistency - we
>never seem to be able to narrow down the cause of the problem.
--
John Hopkin
More information about the Opensim-users
mailing list