I have one server "borrowing" resources on a 32 core Xeon system :biggrin.gif: and another on a VPS - both proper hosting with full 100Mbit connection - not running off my home connection.
I found I was using approx 20KBytes/s of upstream (~160kbit or 0.16Mbit) with 10 players connected to the server, but you will need atleast 2Mbit/s so that people can connect and (a) download the map in a reasonable time (:cool.gif: not lag out the people playing when someone connects.
I couldn't get screen to work with this directly - would just disconnect output - tho you might be able to do screen -dmS minecraft ./caplog.sh I've not tried.
I also have a cronjob doing regular backups but this is kinda cool in that if someone joins a server and does a massive grief you can return it to exactly how it was the moment before they joined rather than hoping you have a close backup.
Been working on a way to backup the server level every time a new player connects... this is still a work in progress and experimental but seems to work...
1. wget http://aten-hosted.com/files/caplog.sh
2. copy to the minecraft folder
3. chmod +x caplog.sh
4. kill any currently running copies of the server
5. start your minecraft ./caplog.sh & (note the & its important)
Now when a player connects you should have a time and dated backup of the server level in the folder minecraft/backup/./.. with the player name at the start.
Should make it easy to keep regular backups and recover from griefing.
When you kill the server process you will need to kill the tail -F nohup.out process as well.
Unfortunatly its not fixed... atleast not for everyone...
While its probably better to wait for Notch to fix this - I solved it on my end with a simple little program - this program actually does NOTHING except enforce a sleep wait... but it should prevent the timer drifting, low framerate and reduce how often you get kicked for "too much lag"... and reduce how bad the fast/slow jerky movement is.
0
0
0
Think you answered your own question.
0
0
I notice my servers crash a lot after awhile if I've got a lot of ports in FIN_WAIT1 status and not clearing.
0
I also have a cronjob doing regular backups but this is kinda cool in that if someone joins a server and does a massive grief you can return it to exactly how it was the moment before they joined rather than hoping you have a close backup.
0
1. wget http://aten-hosted.com/files/caplog.sh
2. copy to the minecraft folder
3. chmod +x caplog.sh
4. kill any currently running copies of the server
5. start your minecraft ./caplog.sh & (note the & its important)
Now when a player connects you should have a time and dated backup of the server level in the folder minecraft/backup/./.. with the player name at the start.
Should make it easy to keep regular backups and recover from griefing.
When you kill the server process you will need to kill the tail -F nohup.out process as well.
0
May or may not work - java seems to be somehow messing with the system timer (which it should never ever do).
My program essentially heads off the deviations before they get too far - but it doesn't completely fix the problem.
0
0
While its probably better to wait for Notch to fix this - I solved it on my end with a simple little program - this program actually does NOTHING except enforce a sleep wait... but it should prevent the timer drifting, low framerate and reduce how often you get kicked for "too much lag"... and reduce how bad the fast/slow jerky movement is.
http://aten-hosted.com/files/mctimerfix.rar
Use at your own risk, etc. - but I promise theres no trojan, viruses, etc. in the file I uploaded for whatever that is worth.