Server Help :: View topic - Reliable packet lost(56) ???

Server Help

Bot Questions - Reliable packet lost(56) ???

AgentSteel - Tue Jun 22, 2004 12:07 am
Post subject: Reliable packet lost(56) ???

Reliable packet lost(56) to: Assault-Bot (pending=735513 current=735768 diff=255)

This shows up in *log 3 or 4 times just before my bot disconnects. Seems to happen with high load, 5k/s or more, and possibly related to when I am actually playing the game while bot is running from my same box.

Any ideas on what this means? I've asked around and have come up with no info sofar.

I noticed this first when I created a statbox that did alot of objon objoff messages... bot didn't last 5 minutes.

Thanks for any help.

CypherJF - Tue Jun 22, 2004 12:28 am
Post subject:

I am not sure, as I am definately not in anyway shape nor form familar w/ the protocol; but is the bot a sysop? I'm not sure if you hit it with a *lowbandwidth 1, will do any good..

Mine GO BOOM - Tue Jun 22, 2004 12:29 am
Post subject:

Taking a guess, it seems as if your bot-server connection lagged enough where there was a backup of reliable packets, so much so that it filled up the buffer (which looks to be 256 packets). Thus, the server could no longer even attempt to buffer reliable packets, so it kicked your bot.

I'd recommend you cut back on how many messages you send. It looks like you are probably trying to send like 500 messages in a second, and your connection cannot handle this. What bot type are you using, like Merv? Also, it isn't hard to add into your bot its own private message buffering system, in which you can make it only send like 10 messages a second, thus to prevent server spamming, and also to prevent loss of syncing between client and server.

AgentSteel - Tue Jun 22, 2004 2:07 am
Post subject:

Thank you for responding.

I have a message buffer that sends message every 100ms... perhaps I should increase that delay???

This is the javacore by Dock>... and bot IS sysop with *bandwidth 30000. Granted... my bot does send alot of messages... but at any one time the max bytes per second is ~5000 and I have a 384/384 dsl.

Hmmm... I'm really stumped. I'm sure MGB is right about the server buffer being filled because I have plenty of available bandwidth, cpu, and memory. I was under the impression that sysop can send unlimited messages... and my messages are even buffered and sent at delayed intervals.

Any ideas on what packet 56 is? It always says 56 in the message, so that is some sort of identifier related to the error.

Mr Ekted - Tue Jun 22, 2004 4:39 am
Post subject:

MGB you are right, except that it's the server trying to send reliable packets to the bot. Maybe the bot is issuing a ton of *info etc. Or it could be that the network is broken S2C or C2S somehow.

Underlord - Tue Jun 22, 2004 10:20 pm
Post subject:

56 is a S2C *watchdamage packet

your bot/connect lagged so much that 256 reliable packets got backed up

isolate the problem by limiting the number of packets being sent/received so fast

CypherJF - Tue Jun 22, 2004 10:39 pm
Post subject:

also try shutting off lowbandwidth then also; since this doubles the packets...

AgentSteel - Thu Jun 24, 2004 12:48 am
Post subject:

Ok, this is the conclusion I've reached.

Since *info *einfo and *watchdamage are all sent as reliable messages... and if you have high enough load... a mere 1000 or 2000ms lag at any one time could be enough to fill up the server's send buffer and boot your bot.

Would have been nice to have *watchdamage be unreliable or another equivilant that was unreliable... or have sysop's be given a larger buffer on the server.

In any case... my solution is to either place the bot on the local network as the server to eliminate the possibility of lag, or simply start turning off bot features which sort of sucks, or spread the features over multiple bots.

Thanks for your help guys.

D1st0rt - Thu Jun 24, 2004 2:18 pm
Post subject:

Quote:

mere 1000 or 2000ms lag

thats quite a bit, 1 or 2 seconds

Most zones won't allow you if you're over 700

AgentSteel - Thu Jun 24, 2004 3:06 pm
Post subject:

Ya.... point is that with large enough load even 700ms could get your bot booted. We aren't talking a constant 700ms ping... but one 700ms ping spike. And given the nature of networking being bursty and spiky and the problem compounding with distance and routers, its not uncommon at all to get a large spike once and awhile.... even once every 24 hours would be annoying.

All bot programmers should be aware of this limit... at some point you will have to disable features, or spread them over multiple bots.

D1st0rt - Thu Jun 24, 2004 3:18 pm
Post subject:

or stop trying to use *watchdamage

Mr Ekted - Thu Jun 24, 2004 3:48 pm
Post subject:

Most zones run their bots on the server itself, or on a machine right next to the server.

AgentSteel - Thu Jun 24, 2004 4:09 pm
Post subject:

Hehe, but *watchdamage is so cool!

I am hoping to get my bot moved to the sscx botserver (the box next to the server ala Ekted

)... but yarkim is so anal about security that I have to work through other people. Bleh.

Maybe the ASSS guys will address this problem, and everyone switches to ASSS, and then... and then... *sigh* oh well.

In any case I'm glad I found this forum, and thanks again for the help in understanding this problem.

(If anyone cares the critical load is about 65 players w/ *watchdamage on with SSS settings. At this point any relatively large ping spike will get your bot booted. I'm also suspicious that since the server is buffering unsent/unacknowleged packets it probably has a timed delay between all sends... and thus, you could have so many packets being sent to you under delay that you will fill up the unsent buffer even w/o a ping spike... something to ponder anyway).

Mr Ekted - Thu Jun 24, 2004 5:00 pm
Post subject:

yarekim controls access to all machines and to the network itself for all of SSCX. He is a really smart guy, and his diligence over security keeps us up and running. He is also a very nice person, but you can imagine how many people are trying to "get things" from him all the time. I can sympathize.

CypherJF - Thu Jun 24, 2004 6:54 pm
Post subject:

He is nice, but is very extremely busy, and normally not get thing's done for like two weeks... :/ but it's alright.. just mostly irritating especially when ya need something lol

Dr Brain - Fri Jun 25, 2004 12:04 am
Post subject:

About switching to ASSS.

Don't do it unless you meet these three requirements:
1.) You need it. This one is very important. Don't move if you want 1 feature it can offer.
2.) You can code. This is a must. You wont find anyone who will do it for you.
3.) Slightly insane. Self explanatory. Usually fufilled by #2.
4.) You have no population to speak of. I made the mistake of not knowing this, and switched while I had a regular population. Only now, 8 months later do I have a population close to what I lost.

If you meet those, you can *think* about moving to ASSS.

In other words, make your bot work for you

AgentSteel - Fri Jun 25, 2004 5:23 pm
Post subject:

Ya, if I were in yar's shoes I would probably as anal or more anal about security than him.

CypherJF - Fri Jun 25, 2004 8:40 pm
Post subject:

If subgame had that one simple command

i'd be happy

so I'm going need to ignore the comment about, "Don't move if you want 1 feature it can offer. "

D1st0rt - Sat Jun 26, 2004 12:46 am
Post subject:

what command is that?

CypherJF - Sat Jun 26, 2004 1:20 am
Post subject:

*warpto but not.. warpto w/ subgame switches the ship's direction... I just need more like *hopto x y .. w/o changing the ships dynamics..

Also being able to tell the server to send all position packets to the bot would help

Smong - Tue Jun 29, 2004 3:39 pm
Post subject:

If you are talking about '?botfeature seeallposn' you might as well just make a module for faster response times.
Since I loaded up linux to figure out why I said what is in d1st0rt's sig. I might as well code the ?moveto command, should be easier than doing ctrl+alt+del then ctrl+f to find that my scroll wheel is messed up and typing a reply closes IE making all topics marked as read -glare at ekted-.

D1st0rt - Tue Jun 29, 2004 10:44 pm
Post subject:

It was because your ftp stuff only worked after the first time