[LONGTERM] Uptime Solution: 'Server Instancing'

Discussion in 'Considered/Other' started by Codemonkey, May 16, 2015.

Thread Status:
Not open for further replies.
  1. Codemonkey

    Codemonkey Registered User Member

    Messages:
    71
    Likes:
    18
    Local Time:
    12:01 PM
    Hello to whom may be reading this,
    Recently, it seems Habboon has (and for a while past that) been suffering numerous downtimes, not for maintenance but as a result of any of the following:
    • Server Crashes (Emulator, VPS, or Content Service)
    • DDoS Attacks (rarely)
    • Mass Disconnection Rates (Where it shows only "11 Users Online!" for example)
    • Other Instances like these.
    These can't really be repaired by normal developer standards, since these are just general things as to how emulators go, and can't be fixed permanently. However, I believe a solution that could work in the long term of Habboon's development would be to start introducing server instancing for when any of the above happen.

    Server instancing, means having multiple copies of the same setup Habboon has (SWFs, DB data, Emulator, etc.) on several servers instead of just the one, so if one server fails, then a script for example, can check uptime and if it sees downtime, can be switched to another server while the main one is being fixed otherwise.

    To make this happen, what could be done is, Habboon developers regularly schedule for perhaps once out the week (during their general maintenance times?) copy all the latest Habboon revision files (emulator/SWFs) to these other servers, and then as for database data, simply (if possible?) have it either be shared amongst all the servers so they all equally write/read the same data, OR copy and upload a recent copy of the database each week to the other servers.

    This is a very, very long term solution I have not yet seen in retros to date, due to it's rather complexity for actual implementation, so I am not asking it be implemented in the next while; in fact, probably not for the next year or two, but given that Habboon has grown to half a million users, it'd be a good solution to combat the above problems, and also to match up to or even surpass Habbo.com's own setup.

    Thanks,
    Codemonkey
     
  2. 123

    123 Registered User Member

    Messages:
    74
    Likes:
    42
    Local Time:
    6:01 PM
    Boon:
  3. 123
  4. I agree this would be a great idea, but for the amount of users we get I don't think its feasible or needed we already outsource at many locations for DDoS protection to help keep us online which seems to doing a great job, and the only downtime we really occur is if we wish to perform updates.

    This is something we'll probably consider in the future once we get larger etc.
     
    Codemonkey and Bren like this.
  5. Codemonkey

    Codemonkey Registered User Member

    Messages:
    71
    Likes:
    18
    Local Time:
    12:01 PM
    Sounds good, given all the factors at the time; hopefully as the game does get larger, it will be considered. Thanks for reading!
     
  6. Sledmore

    Sledmore BoonForums Owner Administrator Forum Owner Member

    Messages:
    306
    Likes:
    692
    Local Time:
    6:01 PM
    Isn't exactly ideal. You're disconnecting because someone is willing to waste money on resources to hit off anything. Our actual only issue is DDoS attacks, I don't think the server has crashed for time.

    You'll see 11 connected users probably due to the proxy being limited on resources due to the fact that some people have nothing better to do, so they'll spend time and money to try and kill anything that is bought.

    I think everything is fine how it is now, each time an attack occurs we can just switch up the proxy and move on.
     
    Adrian likes this.
  7. Codemonkey

    Codemonkey Registered User Member

    Messages:
    71
    Likes:
    18
    Local Time:
    12:01 PM
    Based on what you said, it seems you are only mentioning attacks, right? This solution would apply to more than just DDoS attacks, but rather all the sorts of downtimes; last night for example, the game was inaccessible due to not an attack, but rather just some error producing "The service is not available." or CloudFlare access issues (either one); nobody except a developer was able to fix that until later on when they got online, so this solution would keep the game up while the developers are away, among other things.

    If I read your post wrong, please iterate on what you meant.
     
  8. Sledmore

    Sledmore BoonForums Owner Administrator Forum Owner Member

    Messages:
    306
    Likes:
    692
    Local Time:
    6:01 PM
    That was an attack. Layer 7 flood. Two servers for the website wouldn't fix it, once one server is down and technically "null routed" the traffic would go from A to B and B to A. It wouldn't help, how did we get it back up? I changed the configuration and now it's silky smooth again.
     
  9. Codemonkey

    Codemonkey Registered User Member

    Messages:
    71
    Likes:
    18
    Local Time:
    12:01 PM
    Oh, I see. Well I suppose the solution is not all 'well-rounded' yet, so sorry if it sounds a bit strange in how it was.

    What I was basing the solution off of (in part) for application, was that previously, if Habboon had been down, staff might have told players on the forums to go use one of the other hotels that were (until they shut down) affiliated with Habboon, but replace using another hotel, with instead having players play off a copy of Habboon, with the same (at least, almost; depends on the last database update) data and everything - basically ensuring uptime entirely.
     
Thread Status:
Not open for further replies.