A friggin anooying bug

[Demonic]

I have found a odd bug, you Stick to a spot - normally more if you have a mob (pet) - now i blamed all sorts but another person running on a Dual Zeon with 2 gig ram etc (more than powerfull enough me thinks) is also having it as well as Euromir i think.

The odd thing thing about it - is sometimes you dont get it at all then other days you get it almost all the time - even me local so def not a lag issue.

Has anyone else noticed this who run servers - i tired hexing the server with a few values but the randomness is def still there. One of my thoughts as i say was power of the pc. Now my finding are odd as well, turn on the pc, boot the server and its fine - run the pc for the whole day and the problem is worse - nothing.

One thing I def spot - when the server is running well (not sticking) the little SQL icon flashes away - soon as the problem strikes, - no comms with sql (no flashing Icon on SQL).

Anybody care to add to this?

 

[YouKnowWho]

I had this extact same problem with my version of client/server.

Never did manage to find out why. For me, i belive it was something todo with packets occasinally getting corrupted, or lost.

Is a very strange bug, would be intresting to find out what causes. Any chance you could keep us updated, if you find anything out?

Thanks.

 

[Xander]

Every one gets it i had it on mir 2 aswell its packets getting either corrupt or lost like YouKnowWho said is no way to fix just wait abit then go works for me

 

[Far]

it happens to me everytime i use teleport npc when im on a horse. no idea why

 

[TritoN]

ye same probs on Olymp some times 2...i always used to blame some db underperformance for it but the corrupted packet theorie sounds interesting 2

 

[Demonic]

This is what I found so far:

If i boot the PC (server) and start the server, there is no problems at all for hours on end. If i leave the pc running all day without running server then run server - almost straight away i get the sticky bug. watching the sql icon, the icon is flashing almost all the time but as soon as people stick, sql stops flashing until they move again. CPU Useage all through this process is never much - the chinese have reported a problem where cpu useage goes to 100% for example.

I have tweaked SQl + tried it on a diff machine and still the problem comes about...

 

[konnichiwa]

This is what I found so far:

If i boot the PC (server) and start the server, there is no problems at all for hours on end. If i leave the pc running all day without running server then run server - almost straight away i get the sticky bug. watching the sql icon, the icon is flashing almost all the time but as soon as people stick, sql stops flashing until they move again. CPU Useage all through this process is never much - the chinese have reported a problem where cpu useage goes to 100% for example.

I have tweaked SQl + tried it on a diff machine and still the problem comes about...

every once and a while the server stores ALL character movement data (so i remeber from my old Mir 3 data) and so your slow SQL servers could be causing the lag, as the server does not process this data on a seperate thread.

 

[Mrklawz]

<Yay! back on my revived machine at last>

Corrupt packet transfer seems like the most likely answer. Been messing about on the USA mir3 server and its more pronounced there than any other. Although the ping is fairly high compared with EU mir3, it is constant but it does seem prone to be lossy packet wise. Could be some problem with the CheckSum added to the encrypted packets i guess, Konichiwa might be able to help here possibly.

 

[Demonic]

Well that sounds good but how do you explain a machine freshly booted works fine but a machine running for 8 hours without the server running at all and as soon as the server is run the problem comes, sometimes not.

Running sql on a dual CPU box (1 gig) which is seperate from the Server machine also has the same problems. On this box Im running the server had both 2.9 + SQL and 50-60 people online + God knows how many spawns and map etc never had any problems.

Seen as almost everybody has this problem on all sorts hardware even local login etc means there is a bug in the server code. Euromir3 has this problem also, sure they have enough hardware to chuck at the problem as well. The way to notice the problem more i found is when you have a pet or there is a dead mob in front of you - soon as the dead mob goes, you can run again.

I tested via not having a pet and everything worked a treat, soon as i recalled a mob as a pet, the problem came about. Rebooted PC (server) and even with a pet - no problem. About 4-5 hours later which PC was still running (server) but not the server exe itself - booted server and problem was there in force...

Tweaked SQl as well and although faster at selecting npcs etc - the problem was 100% def there - as its a random problem its hard to pin down what it is to be honest...

 

[konnichiwa]

i would have to assume it comes from either saving the data from memory into SQL, or path finding AI, both are quite expensive, however the first is the most costly

you have SQL running on a "seperate" dual CPU box as in running nothing else apart from sql? what network connection do you have to your server box? how mcuh ram does the database server have?

 

[YouKnowWho]

Well that sounds good but how do you explain a machine freshly booted works fine but a machine running for 8 hours without the server running at all and as soon as the server is run the problem comes, sometimes not.

Well I only mentioned packet lost/corruption, as I know for sure it’s not SQL for me.

The SQL database is accessed very little, logging on and off, and other things to do with items, are basically the only times SQL server is accessed.

The being stuck bug, happens at random points, when none of the above are in action.

 

[Demonic]

i would have to assume it comes from either saving the data from memory into SQL, or path finding AI, both are quite expensive, however the first is the most costly

you have SQL running on a "seperate" dual CPU box as in running nothing else apart from sql? what network connection do you have to your server box? how mcuh ram does the database server have?

Specs for you:

Main Server - Athlon 2500+ barton with 1 gig ram (2 x 512 Dual DDR) 80 GIG HD running off SerialATA (150 speed). Only thing installed is XP Pro SP2, Server and SQl Desktop edition

Other Machine - Dual 933GHz P3 with 1 Gig Ram running Windows 2000 Server fully patched up. Only thing installed on this is SQL Enterprise edition as its role is a server. SQL Tuned up for Dual CPU useage etc..

Connections - 100Mbit LAN and also 400mbit Firewire Lan - all fully tweaked to get 80%+ out of the connection (as LAN normally does 40-50 if not tweaked properly).

Got to remember that its not a live server so most people on at any time is 2, thats me local and another dude from Internet - the problem comes about even if only just me on. When testing alternative SQL machine, i stopped SQL on Athlon Box to free up CPU Power etc and switched all setting to Dual 933.

One thing I have not tired (maybe later) is trying the server fully on Dual 933 but i expect the same problem with come about.

Well I only mentioned packet lost/corruption, as I know for sure it’s not SQL for me.

The SQL database is accessed very little, logging on and off, and other things to do with items, are basically the only times SQL server is accessed.

The being stuck bug, happens at random points, when none of the above are in action.

We do not know to be honest what is causing the bug - I was just reporting when I noticed the problem by watching everything while the other guy was online - so cpu useage, memory, sql chatter etc... Upon the bug happening - SQL does not talk at all - as long as the bug is in effect - how can you be so sure its not your SQL ??

Lets chuck some Variables into the POT so to speak:

1. Hexed Rungate to change time from 10 sec data retension to 1 sec
2. Using 30-06 but i tried even orig exe we first ever had and the same problem (random so hard to check).
3. Mobs been hexed dead time to short time (30 sec i think)

Orig Server never had these things in and still the same problem - nothing reported on Chinese sites at all ref this problem - they as far as I can tell dont have any bugs at all.

 

[NotMeAgainAgain]

Specs for you:

Main Server - Athlon 2500+ barton with 1 gig ram (2 x 512 Dual DDR) 80 GIG HD running off SerialATA (150 speed). Only thing installed is XP Pro SP2, Server and SQl Desktop edition

Other Machine - Dual 933GHz P3 with 1 Gig Ram running Windows 2000 Server fully patched up. Only thing installed on this is SQL Enterprise edition as its role is a server. SQL Tuned up for Dual CPU useage etc..

Connections - 100Mbit LAN and also 400mbit Firewire Lan - all fully tweaked to get 80%+ out of the connection (as LAN normally does 40-50 if not tweaked properly).

Got to remember that its not a live server so most people on at any time is 2, thats me local and another dude from Internet - the problem comes about even if only just me on. When testing alternative SQL machine, i stopped SQL on Athlon Box to free up CPU Power etc and switched all setting to Dual 933.

One thing I have not tired (maybe later) is trying the server fully on Dual 933 but i expect the same problem with come about.



We do not know to be honest what is causing the bug - I was just reporting when I noticed the problem by watching everything while the other guy was online - so cpu useage, memory, sql chatter etc... Upon the bug happening - SQL does not talk at all - as long as the bug is in effect - how can you be so sure its not your SQL ??

Lets chuck some Variables into the POT so to speak:

1. Hexed Rungate to change time from 10 sec data retension to 1 sec
2. Using 30-06 but i tried even orig exe we first ever had and the same problem (random so hard to check).
3. Mobs been hexed dead time to short time (30 sec i think)

Orig Server never had these things in and still the same problem - nothing reported on Chinese sites at all ref this problem - they as far as I can tell dont have any bugs at all.

Hey, its a shame your getting this bug but i have seen no sign of it on my server, i had it running for 4 days constant, between 13 - 20 people playing and no hassles whatsoever, atm i have the sql on the server pc, and i only get the occasional Run DB Wait error, like when im raring or backing up the sql, one bug i do get tho is the undead mobbage, 1/1000 zombies do it, but thats because i had to sacrifice 512mb ram so i can set up a diff pc (It will go back in) usually have 1.5gb in server machine and no hassles whatsoever :S, btw i am using the same set of files that your using...
have you defragged your HDD? :P just a suggestion

 

[buzznett]

I havn't seen this problem as of yet. I switched from the 2004 server to the JOB server and have been running it 24/7 for roughly about 2 months now, only about 15 people play it now while I am working on it.
Now in all fairness I don't let the machine sit without running the server so that may be a factor.
When I update something I do reboot the server exe with no problems even like now, I havn't rebooted the server machine for 3 weeks.

Specs;
Main server is a Compaq Proliant quad zeon 900 w/2megs cach each, 2gigs of ram and the hard drives are ultra320 SCSI 15k running on a raid.

SQL server is a P4 266 with 1gig of pc3400 ram, hard drives are ultra 320 SCSI 10k.

I am also not using a hexed rungate.
I am using the 30-06 exe.

 

[Mrklawz]

Can pretty much say for definate the sticking only happens when what ever is causing the bug allows a player char and a mob/pet to occupy the same space. So its either down to bad collision detection or a missmatch in the update of the coords of each entity. Its not just the JoB files either, its Korean EU and USA server too (i used to put it down to lag on the Korean server, guess i was wrong)

 

[Demonic]

Can pretty much say for definate the sticking only happens when what ever is causing the bug allows a player char and a mob/pet to occupy the same space. So its either down to bad collision detection or a missmatch in the update of the coords of each entity. Its not just the JoB files either, its Korean EU and USA server too (i used to put it down to lag on the Korean server, guess i was wrong)

Yup - as i said on my testing - only happens (if its gonna happen) when i have a pet - its almost like there was something in front of me - as sideways movement is still possible. Just been looking on Chinese forums and I was wrong - they def have the same problem - reffered to on translation as Landing Bug....

 

[Demonic]

Little update:-

I Backed up my Databases from Main Server machine - restored on Dual 933.
Copied over files + orig Rungate to Dual 933.
Ran Server and tweaked SQL while I was at it...

Results so far :

No Sticky bug thing with or without a pet at the moment (though remember i just powered this machine on). Will continue to test at random times leaving machine + server on + get other person to log in much later and see if it comes back at all. Theory been is that a Dual 933 P3 is Slower than a Athlon 2500 Barton (forget Dual CPU - MIR3 Server does not really use it properly). Only Difference i can see, if any is that SQL is Enterprise version where as on Athlon machine its desktop demo + cracked....

We Shall See.....

 

[Demonic]

I havn't seen this problem as of yet. I switched from the 2004 server to the JOB server and have been running it 24/7 for roughly about 2 months now, only about 15 people play it now while I am working on it.
Now in all fairness I don't let the machine sit without running the server so that may be a factor.
When I update something I do reboot the server exe with no problems even like now, I havn't rebooted the server machine for 3 weeks.

Specs;
Main server is a Compaq Proliant quad zeon 900 w/2megs cach each, 2gigs of ram and the hard drives are ultra320 SCSI 15k running on a raid.

SQL server is a P4 266 with 1gig of pc3400 ram, hard drives are ultra 320 SCSI 10k.

I am also not using a hexed rungate.
I am using the 30-06 exe.

I have had a thought - gonna test this right now as it goes....

What Client are you using Buzznet? by which i mean 1.4 / 1.45 / GT feature set. To expand on that in case you dont understand, 1.4 has 2 rows of options and a fixed Spells box. 1.45 has Scolling Spells box but not Green glow on new armours etc - GT been 3 rows of options, Glow on New Armour and scrolling spells box. I have a sneaky feeling the problem is actually client side rather than server at all.... Soon tell you when i get someone to run 1.4 client and test it over a few days...

Quick Note - tested 1.4 client and no mob close but person got stuck for a few seconds. Only thing to note was in line with his pet (horizonlty) then soon as pet moved away (after killing the mob it was attacking) he could move again. Def no other mob near or anything in front of him. This says to me there is a bug in the coding of pets cos remember in 2.9 the pet died as soon as you logged out and got attcked by guards....

 

[NotMeAgainAgain]

This says to me there is a bug in the coding of pets cos remember in 2.9 the pet died as soon as you logged out and got attcked by guards....

EuroEI has the same "Feature".
If you Log out completely your pets die, but if you lag out/hop chars your pet will remain for a set time before dieing, i've allways seen this as a cool bonus

 

[Demonic]

EuroEI has the same "Feature".
If you Log out completely your pets die, but if you lag out/hop chars your pet will remain for a set time before dieing, i've allways seen this as a cool bonus

Im not saying its a bad feature lol - one person is saying (Buzznet) he has never had a bug were people get stuck - we need to work out why, i am pinning my guess on the Pet - without a pet i never get stuck where as with one "sometime" i do and im local to the box so def not lag...I need to elimiate everything and focus our thoughts on what is causing the problem - then and only then can we go somewhere to fixing it.