VPS - Smooth then lag...

[joshtrav]

We tried a VPS some time ago and ruled against it as it just seemed to unstable at higher player counts.

We currently run a 52 slot server through NFO normal hosting options and it runs smooth all the time, no issue, unless a DDoS or some other form of attack.

Our current issue is trying to resolve a good setting for the game server on the VPS. It runs smooth for 2-3 maps with 20ish players, then starts degrading performance wise considerably.

The OS is 2k3, no HPET, we use FireDaemon to run the server as a service, have tried restricting it to a core and not, allocating more or less resources, and have tried numerous config options, this being the most recent fps wise:

Code: Select all

fps_max 200
net_queued_packet_thread 1
sv_parallel_packentities 1
sv_parallel_sendsnapshot 1
sv_max_queries_sec 3
sv_max_queries_sec_global 60
sv_max_queries_window 30
net_splitpacket_maxrate 50000

sv_mincmdrate 20
sv_minupdaterate 20
sv_maxupdaterate 60
sv_maxcmdrate 60
sv_maxrate 50000
sv_minrate 10000

We had tried with and without srcdsfpsboost as well.

At this point we are at a crossroad, we don't want to pay a fortune for 2 game servers when we could pay the same for 1 vps as 1 server pretty much. I read numerous posts by John in other forums saying that NFO's use of VPS can handle game servers even though so many other sites speak of the opposite.

So any help at this point would be appreciated rather then having to bargain shop for a full server somewhere that we have never used, the avenue we very much want to avoid.

[Edge100x]

Our VDSes work very well for game servers. They are configured quite differently than the traditional ones that had performance problems years ago.

You mentioned that your server on the VDS starts off with good performance for the first few maps. Are you saying that restarting the game server or OS fixes your problem? If so, you may be seeing a memory or CPU leak. Does the memory or CPU usage increase for your server over time, as seen in the Windows task manager?

Is the server FPS dipping, as seen in the net_graph 4 output by players?

[joshtrav]

There is no memory leak, we watch it via fire daemon and it holds well within range.

The fps is also reading on with stats. Its almost as if the entire vps just lags behind for a few seconds.

A good example, I cam remote in, and all of the sudden just hang for a good 10 seconds while trying to open and simple txt file. Then boom, as if nothing was wrong it opens and everything is smooth.

It seems directly correlated with the game servers as far as the type of lag. I am just unsure where to venture next.

The fps seems to hold steady around 180ish however, while the ram does increase per client, but holds once we have a steady count, and was well under 200mb the entire time.

[joshtrav]

My answer was really fragmented.

To be more specific:
FPS - holds steady just under 200
Ram - increases per player but stays well within range and under 200mb

The spikes come occasionally during the first 2-3 maps, lasting maybe 3-4 seconds every 10 minutes.

After the 2-3 maps the lag seems almost constant.

While I would venture into leaks its the same exact scripts and setup we are running on our regular hosted NFO server. Down to the same standard configs except for the fps and max client cvars.

The only other thing running on the machine is MYSqL which is monitored and holds at 59mb of ram on most occasions.

[Edge100x]

If you see delays when accessing the disk, it could be that another customer on your physical machine is using it heavily. Turning off logging in TF2 would be one way for you to eliminate this, potentially, but it would be more of a stopgap workaround. Please open a support request and I can check out the disk access from my end in more depth and move you to a new machine if necessary.

[joshtrav]

Sorry about the delay, work has been hectic.

Anyway, here is an example of how easily we can lag out a server.

Boot up SRCDS, have 5 players connected (none really needed), and right click roughly 10 selected files in windows explorer, and boom, the vps lags out for about 10 seconds, and everyone connected to the server looses connection.

In my mind, this should not be even remotely like this. Makes me wonder how overloaded these hard drives are, that a simple access query brings the entire VPS to a halt...

Any ideas on where to go from here?

[Edge100x]

Boot up SRCDS, have 5 players connected (none really needed), and right click roughly 10 selected files in windows explorer, and boom, the vps lags out for about 10 seconds, and everyone connected to the server looses connection.

This would be more about the OS than the VDS, but I can still check the usage on our end, if you'd like. I continue to recommend opening a support request.

You're using remote desktop, and not VNC, correct?

How much memory do you have free? If Windows starts to run out (usually with <300 MB free, it starts behaving poorly), it will start to swap to disk.

In my mind, this should not be even remotely like this. Makes me wonder how overloaded these hard drives are, that a simple access query brings the entire VPS to a halt...

The hard drives are lightly loaded on all our machines, overall. I wanted to check yours specifically just to make sure that it didn't have a temporary usage issue of some sort, as part of the troubleshooting process. It's definitely starting to sound like that's not the cause of your problem, but I'd still like to check it.

[joshtrav]

Using remote desktop.

As far as memory, taskmgr shows 1.91 gig in use, which should leave at least 1.09 gig free. CPU usage resides at roughly 2-8%.

The only things we currently have running are FireDaemon, Daemon, and SRCDS. I will open a ticket as well, but I have no clue what else we could be doing to cause such a problem.