Servers crashed....again

[Chilles]

Uhm, I don't think so. Maybe that is true for game servers but was the server for this forum ever down? I never saw that. Linux servers and meanwhile also Windows servers tend to run pretty stable and most downtimes are scheduled and announced maintanance issues. So if it is only game servers the conclusion is either the enginge (Unity here) or the game itself or both have some major software issues and we all the the now long time nit fixed 3DX bugs. I experienced failing servers but that have been hardware defects.

 

Don't get me wrong: I do not say there are no downtimes, I do not say there are nor failures but I am thinking of the amount of failures.

 

Forum was also Down many times well that's normal. Depending about the needed tasks of each server... A server who use there performance just for a simple Web page or web storage has of course very rare fallouts compared with a game server who use the resources for extensive applications.

 

Server doesn't crash mostly, many times they get stuck in SWAP but the effect is a bad or no connectivity...

The reasons mostly similar specially for Linux Servers!

 

Well it's possible to overload a 10Core and 512GB Ram server with some simple tasks very fast, without the correct maintenance work

[Feuermond]

Forum was also Down many times well that's normal. Depending about the needed tasks of each server... A server who use there performance just for a simple Web page or web storage has of course very rare fallouts compared with a game server who use the resources for extensive applications.

 

Server doesn't crash mostly, many times they get stuck in SWAP but the effect is a bad or no connectivity...

The reasons mostly similar specially for Linux Servers!

 

Well it's possible to overload a 10Core and 512GB Ram server with some simple tasks very fast, without the correct maintenance work

 

Basically Swap means high memory usage. Of course the memory of a server has to be dimensioned adequate but most likely it is memory leaks caused by either improper software implementations which does not free unused memory or it is the memory management in deeper layers which causes the memory to be fragmented and the free gaps between used fragments not to be used any longer.

 

The correct maintenance work is:

a. Be sure your machines run adequate hardware resourcesb.

b. Get your software freeing not used memory

c. Scheduled and announced constant system reboots (temporary solution for b. too, often used to be careless on b. )

 

In contrast waiting until a server crashes is not exactly a professional method. Anyway failures may happen but in this case not having an automatic "server down" alert for 24/7 systems isn't too professional either.

[Chilles]

Linux doesn't erase from self the memory

Also when the application is closed or temporary files aren't needed Linux keeps this in Ram until this moment when the Ram is over and Linux starts to swapping.

This means the system switch files between the Ram in to the Swap file and back... This Task needs extra CPU Resources until the moment who the system needs a Load of 1.00 (100% per Core)...

 

But mostly get stuck the system before

 

The Server owner need set Crontabs and Cronjobs which are based on his system and applications (maintenance work).

Well Linux has many advantages but it does just the things you told the system.

 

Windows cleans the Ram Files sequential from self...

 

 

Everyone can test it by herself:

Use your PC 365days without reboot, use all your programs as always and you will see also windows becomes instability and slow ^^

 

Of course does a Reboot help but this is again a moment (5 - 10min) with Server down and new topics about in Forum

It's sometimes a bit tricky to find and use the best settings for a Linux based Server System...

[Deeborah]

The server stability on a long period view is really good compared with other games.

Having 2 crash in 2 following days seems to me a "never happen before".

 

On an other hand, the nature of the game, (interaction, emotions ....) makes that being afraid to be interupted during a key moment is really tough and can by itself ruin the mood you have when you are in the world.

 

I hear people saying, "do something else that staying here", but when you are in real action, being interupted is very hard to accept.

 

Let's hope a root cause is found and that it is not the effect of the world editor to increase the load of the server and make them crash now more often.

 

Even if said many times, a regular reboot with a fix schedule in the week, looks like a "must have".

[Feuermond]

Of course does a Reboot help but this is again a moment (5 - 10min) with Server down and new topics about in Forum

It's sometimes a bit tricky to find and use the best settings for a Linux based Server System...

What about announced re-boots? Let’s say weekly and every week 6 hours later or so, so everybody will have the bad luck it happens in prime time. The last two server issues here... Sometimes it is about communication with users too, especially with paying users but that is to other threads where I already answered to the communication issue.

[Chilles]

What about announced re-boots? Let’s say weekly and every week 6 hours later or so, so everybody will have the bad luck it happens in prime time. The last two server issues here... Sometimes it is about communication with users too, especially with paying users but that is to other threads where I already answered to the communication issue.

 

Of course is it possible to add automatic reboot sequences and they would very good for hardware maintenance.

Sometimes runs on the server applications which are have a sensitive behaviour about reboots but that's individual dependent...

 

When the fall outs go up compared to planned reboots, well no question which would be the better choice

 

Personal I would search as first a alternate way, for example reduce of useless Server Communication, check the Load at different times of the day, add or edit Crontabs.

 

I think the sync of the User world file for example is useless about two reasons:

They are temporary files who keeps in Ram also when the Designer isn't done with the creation and didn't published the Location. If possible I would just upload this file when the user made the Location public or wants to invite another player. This would be save a lot resources and could also a security step to keep the creations "safe".

 

It's just a guess but I think this many synchronizations of the world files could be a reason for the instability.