Thread

World stops running (Bots)

World stops running // Bots

1  2  |  

ciena

Aug 6, 2003, 10:43am
Does anyone know why when i wipe the world with Xelagot the whole world
poofs off the face of the earth for about 10 min?

builderz

Aug 6, 2003, 1:28pm
Maybe it is due to not enough bandwidth on the host's connection? When a
query or wipe is being done, it takes up a fair amount of bandwidth. If
there are others in the world at the same time the query is going on, it
could "eat up" all of the available bandwidth. Go see my post called
"Remote Admin Tool Transfer Limit" in wishlist for a more detailed
explanation.

Builderz
http://www.3dhost.net

[View Quote]

dlp anne

Aug 6, 2003, 3:10pm
its hosted by AWI so thats not it lol

--
Please include all replies

Anne
Twilyte VR Entertainment
Dreamlandpark
www.dreamlandpark.com
What 3D Virtual Reality Chat is all about!

[View Quote]

jerme

Aug 6, 2003, 3:35pm
This is interesting...

You can cause the world to go offline for about 10 mins by wipeing the world
with an Xelagot?

I have had a long standing problem with the larger world's that I host going
offline for random lengths of time (5-15 mins at a time). The world server
is still running and appears to be fine, but the server stops responding to
client requests, and everyone get the "waiting for server" message.

I always got the idea that this problem was related to having lots of users
in a world at a time. I have expirecned the problem in world's like Broadway
during the New Year's party. But it also occured a lot when I hosted a!!ct,
and there could just be 5 users in the world.

I have never been able to determine *exactly* how to reproduce this bug and
the circumstances required. Thus, the dev deam hasn't been able to help
much. I have made them aware of them problem, but I understand that there's
not much that they can do until the cause is determined.

What is your world's name? dlp anne says your world is hosted by AWI, is
that true? Which Xelagot are you using and what version? What are the
*exact* steps that you have to take from within the Xelagot to reporduce the
bug?

Thanks for your help!

-Jeremy

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jeremy Booker - Owner / Webmaster
JTech Web Systems
www.JTechWebSystems.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Therefore do not worry about tomorrow, for tomorrow will worry about
itself. Each day has enough trouble of its own." -Mathew 5:34




[View Quote]

ciena

Aug 6, 2003, 6:19pm
actually its not hosted by AWI its being hosted by someone else. the path is
hosted by AWI
[View Quote]

baron sweetman

Aug 6, 2003, 7:33pm
"jerme" <JerMe at nc.rr.com> wrote in
news:3f313c56$1 at server1.Activeworlds.com:

> This is interesting...
>
> You can cause the world to go offline for about 10 mins by wipeing the
> world with an Xelagot?
>
If a bot is too 'fast' i.e. sends instructions to the worldserver faster
then it can handle the worldserver times out.

Its a balance between bothost speed, bothost line speed, world load,
worldserver host speed, and wordlserver linespeed which makes it sometimes
appear spurious.

Its easy to reproduce, just setup a bot that sends zillions of messages to
all people in the world and see how fast your server times out....

Just my 2 cents

Baron Sweetman
www.petrossa.com/Chazbot

ciena

Aug 6, 2003, 8:22pm
there was only 3 ppl in a 225x225 world
[View Quote]

ciena

Aug 6, 2003, 11:18pm
oops my mistake the world is hosted with AWI
[View Quote]

andras

Aug 7, 2003, 12:59am
[View Quote] > Does anyone know why when i wipe the world with Xelagot the whole world
> poofs off the face of the earth for about 10 min?
>
>

Wiping the world (erasing the database) causes the world server not to respond for any request for a long time if the database was large. The same can happen when deleting large terrain.
I experimented this problem several times during the last 4 years <I did not have a large enough database before :)>
The same problem happens if a bot issues to much request to the server in a short period of time (creating terrain, mass building, etc).


--
Andras
"It's MY computer" (tm Steve Gibson)

jerme

Aug 7, 2003, 2:17am
Ok, so... now that we have established this as a real, reproducable,
problem.... (and no one can say it's just *my* server that's being weird)...

Shouldn't there be a solution to the problem? Some sort of sytem to throttle
the number of commands the server will accept?

What causes it to go down in the first place? Why can't it handle such a
heavy load? Basicaly, what part of the server breaks when under that much
stress? What determines how long the server is down and how long it takes to
"catch up"?

I imagine the solution is a multi-threaded server model (much like the
Apache model, with the parent control process and children that process
requests), powered by a real database like MySQL.

-Jeremy

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jeremy Booker - Owner / Webmaster
JTech Web Systems
www.JTechWebSystems.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Therefore do not worry about tomorrow, for tomorrow will worry about
itself. Each day has enough trouble of its own." -Mathew 5:34




[View Quote]

bowen

Aug 7, 2003, 3:07am
[View Quote] Files aren't databases, databases aren't files?

--
--Bowen--

No of SETI units returned: 62
Processing time: 51 days, 22 hours.
(Total hours: 1246)
www.setiathome.ssl.berkeley.edu

baron sweetman

Aug 7, 2003, 6:13am
"jerme" <JerMe at nc.rr.com> wrote in
news:3f31d2ec at server1.Activeworlds.com:

> Ok, so... now that we have established this as a real, reproducable,
> problem.... (and no one can say it's just *my* server that's being
> weird)...
>
> Shouldn't there be a solution to the problem? Some sort of sytem to
> throttle the number of commands the server will accept?
>
> What causes it to go down in the first place? Why can't it handle such
> a heavy load? Basicaly, what part of the server breaks when under that
> much stress? What determines how long the server is down and how long
> it takes to "catch up"?
>
> I imagine the solution is a multi-threaded server model (much like the
> Apache model, with the parent control process and children that
> process requests), powered by a real database like MySQL.
>
> -Jeremy
>

The solution is simple, build in waitstates in the bot, preferably
configurable.

The server just gets a buffer overflow and shuts down to prevent damage.
Normal good procedure.

baron sweetman

Aug 7, 2003, 6:16am
"bowen" <Bowen at andras.net> wrote in
news:3f31de7c$1 at server1.Activeworlds.com:

[View Quote] what did Goober say? Oh yes,
http://www.winternet.com/~mikelr/flame21.html ;)

ciena

Aug 7, 2003, 9:01am
thanx everyone :) at least now i know what cause that . its a scary feeling
to think u poofed a whole world! lol
[View Quote]

bowen

Aug 7, 2003, 9:40am
[View Quote] Mm, but aren't they on in the same?

--
--Bowen--

No of SETI units returned: 62
Processing time: 51 days, 22 hours.
(Total hours: 1246)
www.setiathome.ssl.berkeley.edu

jerme

Aug 7, 2003, 12:03pm
I don't think it's that simple. If it's a buffer overflow the whole world
server process/thread *should* die completely. Buffer overflows are fatal
errors, as far as I know.

So, if this is a buffer overflow, why does the process not die and why does
the server wait some (random) period of time before coming back online?

-Jeremy

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jeremy Booker - Owner / Webmaster
JTech Web Systems
www.JTechWebSystems.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Therefore do not worry about tomorrow, for tomorrow will worry about
itself. Each day has enough trouble of its own." -Mathew 5:34




[View Quote]

baron sweetman

Aug 7, 2003, 12:31pm
"bowen" <Bowen at andras.net> wrote in news:3f323aa4$2
at server1.Activeworlds.com:

[View Quote] according to strike rapier not, he says that the dictionary only says a
database is a collection of data. So a memorylist is also a database.... :)

baron sweetman

Aug 7, 2003, 12:33pm
"jerme" <JerMe at nc.rr.com> wrote in
news:3f325c3c$1 at server1.Activeworlds.com:

> I don't think it's that simple. If it's a buffer overflow the whole
> world server process/thread *should* die completely. Buffer overflows
> are fatal errors, as far as I know.
>
> So, if this is a buffer overflow, why does the process not die and why
> does the server wait some (random) period of time before coming back
> online?
>
> -Jeremy
>

i am sure the designer of the wordlserver can answer you much better then i
can. However this is for sure the way i would handle a continuous
overflow.Stop listing for a while and wait for the storm to die down, try
again after a certain period to listen and if the coast is clear start
handling requests again.

just in

Aug 8, 2003, 12:12am
I agree with that procedure - shut down briefly if overloaded... but...

Some hosted worlds do not have this problem occuring - it never happens on
AW's servers and they are managing many worlds with huge traffick loads.
(hmmm maybe it is even occuring on their servers? They do get WFS when
in Alpha World, but it usually is a wait of 20 secs or so - not 5 to 10
minutes).

The problem seems like there still needs a solution, because a local server
should be able to handle the traffick generated by a bot and reasonable
number of users (assuming they are not running on old Commador 64s).

Perhaps there is an overload switch somewhere but its set to an unreasonably
low figure?

~ Justin

[View Quote]

baron sweetman

Aug 8, 2003, 3:41am
"just in" <anon at no.spam..com> wrote in
news:3f3306fb at server1.Activeworlds.com:

> I agree with that procedure - shut down briefly if overloaded...
> but...
>
> Some hosted worlds do not have this problem occuring - it never
> happens on AW's servers and they are managing many worlds with huge
> traffick loads. (hmmm maybe it is even occuring on their servers?
> They do get WFS when in Alpha World, but it usually is a wait of 20
> secs or so - not 5 to 10 minutes).
>
> The problem seems like there still needs a solution, because a local
> server should be able to handle the traffick generated by a bot and
> reasonable number of users (assuming they are not running on old
> Commador 64s).
>
> Perhaps there is an overload switch somewhere but its set to an
> unreasonably low figure?
>
i tend to believe it is a combination of factors. So it will occur on any
where hosted server as long as your bot generates instructions faster
then the server can handle them. It also seems to depend on the time the
server needs to handle each instruction. Setting an attribute is light
load handling message trafic is heavy load I found that too fast
messaging is an effective way to bring down a server.

jerme

Aug 8, 2003, 12:46pm
>I found that too fast
> messaging is an effective way to bring down a server.

That would explain why 50 people at a New Year's Eve party could bring down
the server, especially if they're all talking at once.

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jeremy Booker - Owner / Webmaster
JTech Web Systems
www.JTechWebSystems.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Therefore do not worry about tomorrow, for tomorrow will worry about
itself. Each day has enough trouble of its own." -Mathew 5:34




[View Quote]

andras

Aug 8, 2003, 1:20pm
[View Quote] >
>
> That would explain why 50 people at a New Year's Eve party could bring down
> the server, especially if they're all talking at once.
>

naww - that wouldn't bring the server down - maybe congest the pipe it is connected. I had 80 users clicking and chatting like crazy for 36 hours on a P2 300MHz with not a single hickup.
OTOH I can bring down the server with a single bot (if it is on the LAN!) simply building too fast.
The problem lies on the database management IMHO.

--
Andras
"It's MY computer" (tm Steve Gibson)

baron sweetman

Aug 8, 2003, 1:43pm
"andras" <andras at andras.net> wrote in news:3f33bfdb$2
at server1.Activeworlds.com:

>
>

mmm, i had a bot which upon logging into the world send everyone a message
about the whereabouts of everyone else (don't ask) everytime it logged in
the server shut down

I was not building at all

kf

Aug 8, 2003, 3:33pm
I only changed, manually and several times, multiple world attributes in
a world with 15 people in it and the world server crashed. :-(



[View Quote]

jerme

Aug 8, 2003, 5:28pm
> naww - that wouldn't bring the server down - maybe congest the pipe it is
connected. I had 80 users clicking and chatting like crazy for 36 hours on
a P2 300MHz with not a single hickup.

That's odd. My server is a Dell PowerEdge PIII 800Mhz (running Red Hat Linux
7.2, what OS are you using?). So if that's the case then it should have
plenty of processing power.

The server is also co-located in a data center. It shares a 5Mb/s line with
3-5 other systems in the same rack. However, the Cisco switch there is
carefully monitered. If anyone should be hogging the line (and thereby
cutting off service to my server) the people in the control center will know
about it right away. So, I doubt bandwidth is the problem here.

> OTOH I can bring down the server with a single bot (if it is on the LAN!)
simply building too fast.

What bot do you use to do this? Approximitly how fast do you have to build?
I'd like to be able to reproduce this with a linux test machine I have on my
LAN.

> The problem lies on the database management IMHO.

I whole-heartedly agree. The world server design should be moved to a
multi-threaded model. If done properly, this would allow for much higher
loads. (Though I realize the problems this raises). Each child process then
could become a client to a full-blown RDBMS (like MySQL, for example).

A database system like MySQL can handle an *insane* number of queries. It
would also make administration much simpler. (i.e. issue SELECT, DELETE,
INSERT statements directly to the database, rather than having to worry with
the world server.)

This would open up the doors to many neat web applications, with PHP, PERL
and other programing languages that can interface with MySQL.

-Jeremy


--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jeremy Booker - Owner / Webmaster
JTech Web Systems
www.JTechWebSystems.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Therefore do not worry about tomorrow, for tomorrow will worry about
itself. Each day has enough trouble of its own." -Mathew 5:34




[View Quote]

baron sweetman

Aug 8, 2003, 6:54pm
"jerme" <JerMe at nc.rr.com> wrote in
news:3f33f9f9$1 at server1.Activeworlds.com:

> connected. I had 80 users clicking and chatting like crazy for 36
> hours on a P2 300MHz with not a single hickup.
>
> That's odd. My server is a Dell PowerEdge PIII 800Mhz (running Red Hat
> Linux 7.2, what OS are you using?). So if that's the case then it
> should have plenty of processing power.
>
> The server is also co-located in a data center. It shares a 5Mb/s line
> with 3-5 other systems in the same rack. However, the Cisco switch
> there is carefully monitered. If anyone should be hogging the line
> (and thereby cutting off service to my server) the people in the
> control center will know about it right away. So, I doubt bandwidth is
> the problem here.
>
> simply building too fast.
>
> What bot do you use to do this? Approximitly how fast do you have to
> build? I'd like to be able to reproduce this with a linux test machine
> I have on my LAN.
>
>
> I whole-heartedly agree. The world server design should be moved to a
> multi-threaded model. If done properly, this would allow for much
> higher loads. (Though I realize the problems this raises). Each child
> process then could become a client to a full-blown RDBMS (like MySQL,
> for example).
>
> A database system like MySQL can handle an *insane* number of queries.
> It would also make administration much simpler. (i.e. issue SELECT,
> DELETE, INSERT statements directly to the database, rather than having
> to worry with the world server.)
>
> This would open up the doors to many neat web applications, with PHP,
> PERL and other programing languages that can interface with MySQL.
>
> -Jeremy
>
>

if memory serves me well at least the unverse servers can run on top of
MySQL, it seems logical worldservers do the same. But thats speculation i
have no knowledge of this.

And if the worlddatabase was the issue here, how come you can bring down
a world by merely sending a lot of messages? The citizens are kept in the
universe server, so the worldserver doesn't lookup any names in the
database.

andras

Aug 8, 2003, 7:32pm
[View Quote] <snip>
> That's odd. My server is a Dell PowerEdge PIII 800Mhz (running Red Hat Linux
> 7.2, what OS are you using?). So if that's the case then it should have
> plenty of processing power.

Well,,,, I had problem with the Linux version, so I run that particular event on a w2k machine. Linux crashed the same way than the original post claimed, so I just switched to the windows version.

>
> What bot do you use to do this? Approximitly how fast do you have to build?
> I'd like to be able to reproduce this with a linux test machine I have on my
> LAN.

I use DEM2Rwx bot to create terrain (LARGE ones!) and sometimes when I used it to build (or use YASBB to build) it caused the same problem.
They can be found on my website: http://www.andras.net

>
>
> This would open up the doors to many neat web applications, with PHP, PERL
> and other programing languages that can interface with MySQL.
>

That would be nice but the server uses C-Tree for database handling (and that reflects to Baron Sweetman's post too).

> -Jeremy
>
>


--
Andras
"It's MY computer" (tm Steve Gibson)

baron sweetman

Aug 8, 2003, 7:44pm
>
> That would be nice but the server uses C-Tree for database handling
> (and that reflects to Baron Sweetman's post too).
>

tnx Andras. Seems like a weird bunch of hardly connected programs though.
The uniserver runs on ini's or SQL, the world runs on Ctree, the SDK
queries both....Cant say this looks like a well planned enterprise :)

Still doesnt retract from my experience that sending a lot of messages
brings down the world, thereby showing that its not a database problem
but a buffer problem.

jerme

Aug 8, 2003, 8:38pm
> And if the worlddatabase was the issue here, how come you can bring down
> a world by merely sending a lot of messages?

It could be either... Whether it's building commands or chat events, it's
all the same stuff. Sending enough commands fast enough will kill the
server.

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jeremy Booker - Owner / Webmaster
JTech Web Systems
www.JTechWebSystems.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Therefore do not worry about tomorrow, for tomorrow will worry about
itself. Each day has enough trouble of its own." -Mathew 5:34




[View Quote]

just in

Aug 9, 2003, 2:20am
At the new years events and 4th July events we do have multiple bots
running... one for entry and exit monitering and the other for special
effects (fireworks and countdown). The bots are run on a different server
but they do account for some messaging and for building and editing of
objects. I can't recall if the downtime is coincedent with the effects
which occur on the hour.

Regards, Justin


[View Quote]

1  2  |  
Awportals.com is a privately held community resource website dedicated to Active Worlds.
Copyright (c) Mark Randall 2006 - 2024. All Rights Reserved.
Awportals.com   ·   ProLibraries Live   ·   Twitter   ·   LinkedIn