T:RP Instability
- Thread starter posepsi
- Start date
- Watchers 0
- Status
Nicco
danger
- Joined
- May 21, 2011
- Messages
- 0
- Reaction score
- 41
Just open task manager and look at the performance tab during peak times as Arzalis said. This will prove straight away if the box is struggling with the amount of servers it's having to run. If you want a more indepth view on things then go to start -> administration tools -> Peformance monitor
If the server has a single CPU you seriously need to upgrade, how much RAM does the box have? I'm certain that if you've stuck with the same provider for all these years, you may be able to get some kind of a discount if you want to 'upgrade' the box.
I'm also sure that quite a few people would chip in for the extra cost.
If the server has a single CPU you seriously need to upgrade, how much RAM does the box have? I'm certain that if you've stuck with the same provider for all these years, you may be able to get some kind of a discount if you want to 'upgrade' the box.
I'm also sure that quite a few people would chip in for the extra cost.
pug johnson
Has absolutely no life
- Joined
- May 22, 2011
- Messages
- 759
- Reaction score
- 325
hueheueheu
- Joined
- May 16, 2011
- Messages
- 3,845
- Reaction score
- 4,886
Thus I appreciate it when people approach things decently instead of wailing "TNB BOX IS SHIT, UPGRADE" when they have no idea. Im sure people can understand my reserved nature when im being asked to update the box based on hearsay in ventrilo.
Only last week I was speaking to the hosts, well before most of this thread took place, and got a discount on the current bill. However obviously it is more important to have a box that performs, rather than being cheap, we can all agree.
At the time of writing this thread I have 6 server IPs online, running 4 x Tacoscript and 2 x Tacoscript2. Granted, not many people are online right now, so Ill take that into consideration. The CPU is running at 20%. A guestimate would suggest at peak times (from memory) the CPU runs at around 50%. Ill have to double check that though. The only time I remember the CPU really struggling was during a weak where Advantage tried out an auto-restart feature/program which ate up loads of cpu for some reason.
Problems I can see?
- Custom models, yes. Take into account however that my method of minimalistic modelling means that 90% of the models Ive made in the last 6 months all refer to the same few texture sheets, apart from the faces.
- Tacoscript1s errors. For some reason we've never managed to fix certain issues with TS1, and the console is constantly spamming errors. Just looking at the DOS window, you cant see anything but the same spammed error, continously. Weve been ignoring due to the fact we're switching to TS2 within a week hopefully.
- Box outdated? Its not the most amazingly fresh box, no. We could, and do intend, to upgrade.
However I urge you not to take this idea of a "new host" lightly, as it is a big change, and a LOT of work for me and the other RDAs etc. The amount of times we've had to completely wipe the box due to some retarded thing with stan, pac or whatever, has already shown me that setting up is a long winded and severely unpleasant task that I wish to avoid wherever possible.
I also want to make sure that If we do take the path of a new host, we dont end up with some shit company who are going to be worse than the one we have now.
Only last week I was speaking to the hosts, well before most of this thread took place, and got a discount on the current bill. However obviously it is more important to have a box that performs, rather than being cheap, we can all agree.
At the time of writing this thread I have 6 server IPs online, running 4 x Tacoscript and 2 x Tacoscript2. Granted, not many people are online right now, so Ill take that into consideration. The CPU is running at 20%. A guestimate would suggest at peak times (from memory) the CPU runs at around 50%. Ill have to double check that though. The only time I remember the CPU really struggling was during a weak where Advantage tried out an auto-restart feature/program which ate up loads of cpu for some reason.
Problems I can see?
- Custom models, yes. Take into account however that my method of minimalistic modelling means that 90% of the models Ive made in the last 6 months all refer to the same few texture sheets, apart from the faces.
- Tacoscript1s errors. For some reason we've never managed to fix certain issues with TS1, and the console is constantly spamming errors. Just looking at the DOS window, you cant see anything but the same spammed error, continously. Weve been ignoring due to the fact we're switching to TS2 within a week hopefully.
- Box outdated? Its not the most amazingly fresh box, no. We could, and do intend, to upgrade.
However I urge you not to take this idea of a "new host" lightly, as it is a big change, and a LOT of work for me and the other RDAs etc. The amount of times we've had to completely wipe the box due to some retarded thing with stan, pac or whatever, has already shown me that setting up is a long winded and severely unpleasant task that I wish to avoid wherever possible.
I also want to make sure that If we do take the path of a new host, we dont end up with some shit company who are going to be worse than the one we have now.
I'll post a bigger response when I get home, but textures aren't loaded server side and models are only used for physics calculations.
Right. Basically, it's not only player entities on the server that use the CPU core. In fact, they use a very very small amount. It's the physics models for each prop, player model, npc model, hitbox, and physics calculations. What this means is you can't say "Oh, there's 5 people on right now and it's using 20% of the core, so 10 people will use 40%." The amount of cpu that's used isn't additive or multiplicative, it's exponential and has way too many factors to be calculated so easily. The reason the server usually dies during some kind of a fire fight is because the CPU has to handle every single one of those hit box calculations and also do the math to determine how much damage is done, etc.
To further explain what I said earlier, texture are not loaded on the server at all. People think combine props cause more lag because of the nifty little effect they have, but to the server they don't. The only thing the server is concerned about in the physics model (which I'm sure you've seen how simple those look in phy.smd when you decompile a model, Dave.) The only reason you even have texture files on the server is so they're in the fast download. Try a test for me, if you want. Watch the cpu usage then take two ragdolls and start hitting them together and look at the CPU usage again. If it spikes more than just a few percent there's an issue.
On the note of transferring between servers, it's actually very easy if you do it with some thought. You pay for a server ahead of time, usually, so what you do is get the new server about a week before the current one actually expires. You can then use FTP and transfer the majority of stuff between the two servers directly. With a 100 mbps+ line (standard) it won't take long at all. SRCDS doesn't need registry entries or anything to work properly. You can literally drag and drop. The only thing you'd need to install really is MySQL and then you import the tables on the new server (again, very fast due to connection speed.) There are obviously small things that can arise, but it shouldn't take longer than an hour or so to get everything running again on a new server after everything is transferred over.
To further explain what I said earlier, texture are not loaded on the server at all. People think combine props cause more lag because of the nifty little effect they have, but to the server they don't. The only thing the server is concerned about in the physics model (which I'm sure you've seen how simple those look in phy.smd when you decompile a model, Dave.) The only reason you even have texture files on the server is so they're in the fast download. Try a test for me, if you want. Watch the cpu usage then take two ragdolls and start hitting them together and look at the CPU usage again. If it spikes more than just a few percent there's an issue.
On the note of transferring between servers, it's actually very easy if you do it with some thought. You pay for a server ahead of time, usually, so what you do is get the new server about a week before the current one actually expires. You can then use FTP and transfer the majority of stuff between the two servers directly. With a 100 mbps+ line (standard) it won't take long at all. SRCDS doesn't need registry entries or anything to work properly. You can literally drag and drop. The only thing you'd need to install really is MySQL and then you import the tables on the new server (again, very fast due to connection speed.) There are obviously small things that can arise, but it shouldn't take longer than an hour or so to get everything running again on a new server after everything is transferred over.
- Joined
- May 16, 2011
- Messages
- 3,845
- Reaction score
- 4,886
Arzalis said:
Most of that I already assumed / thought i knew, but yes its good to get a bit of clarity.
Like I said, from the top of my head, CPU usually hits 60% at most during peak times. Ill try and check when the servers get packed again next.
HL2RP does not crash anywhere near as much as TRP. It also does not suffer from the periodic lag bumps that seem to be almost timed. (watch TRP on HLSW during above 30 player times, and you can actually see these regulated spikes.)
I always assumed it would be a lag spike due to a time-delay weapon saving command or something, but never actually found out.
Try giving more RAM to MYSQL, see if that fixes the lag spurts. I know on SRP (when it was here) I noticed it'd lag literally every time someone joined and chose a character. That went away at GGL, but we also let MySQL go crazy with RAM usage instead of impose a limit on it. It'd definitely be something to test to try and rule out.
I know the RAM usage from SQL would sometimes go up to 4 GB (We had 16 GB,) but we also ran the entire website off that box too. So I'd wager 75% or so of that was from VBulletin.
I'm also working blind here so it's hard to say exactly what things could be. I'm still convinced the majority of the issues are simply because of dated hardware. If you ever were looking for a new provider, I can recommend a few. There's one in particular (JestServers) who is beyond exceptional when it comes to support.
I know the RAM usage from SQL would sometimes go up to 4 GB (We had 16 GB,) but we also ran the entire website off that box too. So I'd wager 75% or so of that was from VBulletin.
I'm also working blind here so it's hard to say exactly what things could be. I'm still convinced the majority of the issues are simply because of dated hardware. If you ever were looking for a new provider, I can recommend a few. There's one in particular (JestServers) who is beyond exceptional when it comes to support.
smt
Member
- Joined
- Jun 5, 2011
- Messages
- 23
- Reaction score
- 1
"Someone give me proof/informed guess of it actually being anything to do with the box and Ill take it seriously."
"TnB is a Pentium 4 with 4 GB of RAM, I believe?"
If the latter is correct (if it's a troll I'll look dumb but hey, in case it is true...), I don't see why you're asking for a reason. Pentium 4s are incredibly old, and in terms of hardware, even if your box had the latest pentium 4 (2008) that is incredibly old, and 4 gigs of ddr2 (i presume, with an older cpu) is what I have in my home PC, that is no where near capable of running 5 or 6 servers, I don't know how it managed it before.
For arguably the largest serious RPing community, the box should be pretty top end (i7/xeon) and 8 or so gigs of ram. Hardware depreciates over time, if this old hardware has been running for 2 years, I honestly don't see how you don't see a reason.
"TnB is a Pentium 4 with 4 GB of RAM, I believe?"
If the latter is correct (if it's a troll I'll look dumb but hey, in case it is true...), I don't see why you're asking for a reason. Pentium 4s are incredibly old, and in terms of hardware, even if your box had the latest pentium 4 (2008) that is incredibly old, and 4 gigs of ddr2 (i presume, with an older cpu) is what I have in my home PC, that is no where near capable of running 5 or 6 servers, I don't know how it managed it before.
For arguably the largest serious RPing community, the box should be pretty top end (i7/xeon) and 8 or so gigs of ram. Hardware depreciates over time, if this old hardware has been running for 2 years, I honestly don't see how you don't see a reason.
I'm not entirely sure if it's factually correct, but it wasn't a troll. Was told that by one of the higher ups a while back (around the time when I was still an admin.)
smt
Member
- Joined
- Jun 5, 2011
- Messages
- 23
- Reaction score
- 1
Pentium 4s aren't even dual core, sure they're hyper threaded but they're desktop CPUs :( I hope that's false/changed now though, but It'd be hard not to mention the amount of models and materials loaded when someone connects to the server. I don't know the technical specifications of Elysium's old box, but the script ran like a fucking dream on there, I had none of these lag spikes and could join without crashing too.
They're not even Desktop CPUs anymore. They're just old.
- Status