Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


After a LEB to *reliably!* run SRCDS, or to just get SRCDS running reliably on my current one
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

After a LEB to *reliably!* run SRCDS, or to just get SRCDS running reliably on my current one

RophRoph Member
edited January 2013 in Help

For those unaware, SRCDS is Source Dedicated Server - the server daemon for all of Valve's multiplayer games. I'm looking to exclusively run a TF2 server on it.

Now what do I mean by reliably?
Well, I'm having rotten luck trying to run srcds on a 2GB ChicagoVPS box. Yeah, 2GB. It never goes near the 2GB limit, but the server process is constantly going down / being suspended. I'm still not sure what's going on here or why this happens, though I'm assuming it's something to do with OpenVZ and/or Overselling.

ChicagoVPS support moved me onto a new node where I was the only customer, and it ran perfectly for days. Then it started to happen again, I'm guessing as new customers were added to the node. A LEB I have solely for running SRCDS on isn't much use to me if SRCDS is constantly going down. I'm interested in input from some of you guys as to what might be stopping SRCDS and how I could avoid it?

This isn't a dig at ChicagoVPS - at $30/yr for the 2GB VPS I'm not expecting diamonds. And I'm well impressed with CVPS support throughout this, they've gone above my expectations. But I'm still stuck with it not really working. Unless there's some mystery fix, I'm looking for a new LEB where SRCDS will run and stay running.

Are there any SRCDS gurus out there? I'd like to hear how you have it running, what kind of specs you're running it on, especially what virtualisation type.

If I'm going to be getting a new LEB, the specs I'd expect:

768-1.5GB RAM

~15GB space - The game itself takes about 9 gigs, and it's constantly having new content added, so I'd like some breathing room for that and logs etc.

I'd prefer fewer, more powerful virtual CPU cores over multiple weaker cores.

500GB transfer should be enough, I'd prefer latency over throughput but neither are too important, it's not a serious server.

Budget: $3-6/mo or equivalent - I'd be interested in 4/6/12 month offers with the appropriate discounts.

«134

Comments

  • NHRoelNHRoel Member
    edited January 2013

    @Roph said: 768-1.5GB RAM

    This ^

    @Roph said: $2-4/mo

    and this ^

    for

    @Roph said: Budget: $2-4/mo

    This budget^

    @Roph said: Overselling.

    Is only possible by doing this ^ and

    @Roph said: , but the server process is constantly going down / being suspended.

    You will face this again and again with any provider.

    Increase your budget a little bit :)

  • @NHRoel said: @Roph said: , but the server process is constantly going down / being suspended.

    You will face this again and again with any provider.

    Increase your budget a little bit :)

    I don't understand why the process would go down or be suspended. Does the process kill itself if it can't allocate some X % of CPU? Do you need dedi cpu?

  • jarjar Patron Provider, Top Host, Veteran
    edited January 2013

    I have never had an issue with people running this. I have several clients doing it and it runs just fine. Sure it bursts high every now & then but that's about it. It's the least abusive game server I've ever seen.

  • @Corey said: I don't understand why the process would go down or be suspended. Does the process kill itself if it can't allocate some X % of CPU? Do you need dedi cpu?

    System is running out of memory most likely and thus openvz doing what it does best for memory management, killing processes.

  • @Corey said: I don't understand why the process would go down or be suspended. Does the process kill itself if it can't allocate some X % of CPU?

    From the perspective of a player / game client, the server goes down. If you're playing when it happens, everything stops and you timeout. If you try and reconnect, you can't.

    Looking at top, the server process is still running and still eating roughly the same amount of memory as before it went down. It will never do anything again though until you restart it. It just becomes a zombie.

  • @NHRoel said: System is running out of memory most likely and thus openvz doing what it does best for memory management, killing processes.

    Overselling a node does not mean you let the node run out of memory.....

  • I have run TF2 reliably in my 256mb ovz plans ($4) so it is even in you budget.

  • Here's an example, it's "down" now:

    image

    The only way it will ever work again is if I restart srcds.

  • @Corey said: Overselling a node does not mean you let the node run out of memory.....

    Looking at the budget/price for 2gb vps on e3 node op is paying, you have to be really dumb to believe that host is not badly overselling memory and a hosts worst nightmare, customer starts using what they paid for @jarland :P.

  • @Roph said: The only way it will ever work again is if I restart srcds.

    Aything ringing up massive amounts of fail counts in /proc/user_beancounters?

  • jarjar Patron Provider, Top Host, Veteran

    @Roph said: Here's an example, it's "down" now:

    Perhaps limiting CPU on the process to unusable levels with cpulimit? I do it to abusive processes. Never done it to that one because it rarely bursts over 30% that I see.

  • jarjar Patron Provider, Top Host, Veteran
    edited January 2013

    @NHRoel said: customer starts using what they paid for @jarland :P.

    Hell will freeze over first ;)

    I think I've sold around 40-50GB on my current 64GB Dallas node though. They can use all they want, I'm not selling much more on it because it's a single E5. Got a dual E5 coming between tomorrow and Friday.

  • exussumexussum Member
    edited January 2013

    i says have issues with games like this so i have a cron job restart the server every day at a quiet point.

    kill the Linux process and the run process will respawn. i currently don't check for active users but you could add that check

  • RophRoph Member
    edited January 2013

    @miTgiB said: Aything ringing up massive amounts of fail counts in /proc/user_beancounters?

    edit: or a link to the image, pretty wide

  • AnthonySmithAnthonySmith Member, Patron Provider

    Does it have to be USA based?

    OpenITC has a good KVM offer on LEB at least that way you can actually use the RAM your paying for, little more expensive though.

  • @AnthonySmith said: actually use the RAM your paying for, little more expensive though.

    Isn't that the way it is supposed to be :)

  • actually thinking of it have there been any updates to the game ? some games won't let old server versions connect to new game versions.

    killing the Linux process as above will also check for updates and install them automatically

  • AnthonySmithAnthonySmith Member, Patron Provider
    edited January 2013

    @NHRoel said: Isn't that the way it is supposed to be :)

    Indeed :)

    I don't want to annoy/upset anyone but when you pay $0.0012 p/MB p/month equivalent (or the equivalent of 1 cent p/8MB) for 2GB Ram its not a huge surprise that its sadly not the way it is, sadder still is the acceptance of this practice.

  • MelitaMelita Member, Host Rep
    edited January 2013

    I've read your topic on srcds forum, and it seems that VPS is not a problem when you still see srcds_linux process alive. You're trying to run TF2 idle server? I've encountered this same problem. Out of 45 TF2 server running, sometimes 1 ~ 2 of them will randomly locked up almost every days (the player in game will just having timeout and disconnected. Top / ps commands still showing srcds_linux as alive (means it never crash, just lockup). This circumstance happened more often with idle and MvM server.

    When this is happened (I monitored it using HLSW, checking every morning), what I do is just killing off that process manually, so the -autoupdate script kicks off and starts the server again. I was thinking this problem is related with the TF2 itself (it mentioned many times in valve mailing list), because I never found this happened with my csgo and l4d2 server.

    Oversold nodes can makes your srcds_linux to be randomly killed if it exceeds your guaranteed RAM (that's why I prefer VSwap OpenVZ), but I think that's not the case with you, since you still seeing srcds_linux alive. And OpenVZ is the right choice on hosting srcds (provided your nodes isn't oversold), because Xen HVM & KVM introduces some kernel latency (can be seen in net_graph 4) and CPU usage penalty.

  • •1536 MB RAM
    •3 vCPU Cores
    •60 GB Raid10 Diskspace
    •750 GB Bandwidth
    •1 IP
    »»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»»
    $5.25 p/m - Monthly, Quarterly, Bi-annual, Annual
    https://billing.24khost.com/cart.php?a=add&pid=146

    Again it is openvz but that is our offer that is in the offer section. Memory upgrades are cheap if needed.

  • herbyscrubherbyscrub Member
    edited January 2013

    I'm actually running CS:GO 32 man on a CHVPS 2GB without problems. It's only full for a couple hours each night though, but I haven't run into any problems during that time.

    Debian Squeeze 32 bit with Sourcemod.

  • @NHRoel said: @Roph said: 768-1.5GB RAM

    This ^

    @Roph said: $2-4/mo

    and this ^

    for

    @Roph said: Budget: $2-4/mo

    This budget^

    @Roph said: Overselling.

    Is only possible by doing this ^ and

    @Roph said: , but the server process is constantly going down / being suspended.

    You will face this again and again with any provider.

    Increase your budget a little bit :)

    ^
    @NHRoel Made me cry :( It was the most beautiful thing i ever saw anyone do on LEB/LET. Bravo Sir.

  • @Roph said: Well, I'm having rotten luck trying to run srcds on a 2GB ChicagoVPS box. Yeah, 2GB. It never goes near the 2GB limit, but the server process is constantly going down / being suspended. I'm still not sure what's going on here or why this happens, though I'm assuming it's something to do with OpenVZ and/or Overselling.

    ChicagoVPS support moved me onto a new node where I was the only customer, and it ran perfectly for days. Then it started to happen again, I'm guessing as new customers were added to the node. A LEB I have solely for running SRCDS on isn't much use to me if SRCDS is constantly going down. I'm interested in input from some of you guys as to what might be stopping SRCDS and how I could avoid it?

    This isn't a dig at ChicagoVPS - at $30/yr for the 2GB VPS I'm not expecting diamonds. And I'm well impressed with CVPS support throughout this, they've gone above my expectations. But I'm still stuck with it not really working. Unless there's some mystery fix, I'm looking for a new LEB where SRCDS will run and stay running.

    Roph,

    That is incorrect, the node you are on is still empty and remains in "locked" status, we never got around to unlocking it to be used for new orders as we agreed to keep the node empty for you for a while to evaluate SRCDS.

    The node is still empty and I'm sure any benchmark you run on the server can prove this... so overloading is out of the question here.

    This is a 3rd party issue with your SRCDS config, or something of that nature, @Melita has provided some excellent insight that might help you get pointed towards the right direction as I personally am not a "gameserver tech" by any means.

    Thanks,

    Kevin

  • If from the perspective of client the server went down, it might be the CPU or Memory.

    Used to run 3 L4D2 on OpenVZ with 1.5GB/2 Cores. Now, I am on a KVM.

  • mojedamojeda Member
    edited January 2013

    I've ran at least 2 full TF2 32 slot servers on a 2 GB ChicagoVPS with no problem, previously. RAM usage never really went above 1GB, if it did it was barely.

  • How much does the SRCDS RAM usage differ between the different source games? I've been running 32 slot servers on 256mb RAM

  • Maybe try running it without the installed SRCDS? I didn't even know that method existed, I've always used:

    ./hlds/server/orangebox/srcds_run -game tf -autoupdate -steambin /home/tf2one/hlds/steam -port 27015 -maxplayers 32 +map ctf_2fort

    Just make sure to update the file paths as needed.

  • BlazeMuisBlazeMuis Member
    edited January 2013

    Increase your budget a bit as others also said
    SRCDS uses the CPU and the Disk alot, get a XEN or KVM VPS (With decent specs) and you wont have this problem anymore

    Oh and to update you need to use: ./steam -command update -game tf2 -dir (Your dir here) i assume you used the .hldsupdatetool to get the server files

  • @joodle said: get a XEN or KVM VPS (With decent specs) and you wont have this problem anymore

    I don't see how getting a XEN or KVM VPS will solve the issue of CPU.

Sign In or Register to comment.