Howdy, Stranger!

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


ServerBear script causing heavy IO and got my VPS suspended - Page 2
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.

ServerBear script causing heavy IO and got my VPS suspended

2»

Comments

  • jarjar Patron Provider, Top Host, Veteran
    edited September 2012

    @fileMEDIA So do the benchmarks return low or does no one else on the node get to burst resources during that time? You can manage but you can't be running hardcore IO benchmarks and act like your other clients can do disk intensive operations at the same time without either feeling a bit of a loss in average performance.

  • ChanChan Member
    edited September 2012

    Quick Update: Edis just re-enabled my VRS and everything is fine again. I don't think that it's their fault after all as my VRS did do heavy IO for like 30 minutes.....

    @fileMEDIA: You're right in a way but Edis is selling a "fair-share" of resources for the VRSes, instead of "guaranteed" which they do with the KVMs. Which in a way is better for the customers as they can technically use all that's available from the server host as long as no one else is using it, but not so good when there are abusers

  • acarrascopeacarrascope Banned
    edited September 2012

    Is reasonable think in a review to @serverbear script - "several hours" - must be annoying for nodes admins. It is a great value to all us but sure I/O test are exhausting

    pd : Madrid 0 - 1 City -> draw 1 - 1

  • @24khost said: Did everybody wake up and have somebody piss in there wheaties?

    Yes, @Randy

  • @jarland Limit the process with ionice, dm-ioband or something other..No problem here..All communication to the io layer are taking by a process (eg. xen: blkback.xxx.xvda), you can limit it with all working tools.

    Onapp included this function, too!

    I think it´s not a good business if you suspend a vps when it using some io ressources in hour or so.

  • Sorry to hear you got suspended, generally the script takes ~40-50 minute to run. If you've got more CPU cores it takes longer, from our testing if shouldn't place any unnecessary stress on other users on the node unless:

    The node performs poorly - the tests will run longer (poor IO speed/download speed will dramatically increase the length of those tests)
    Multiple people are running it on the same node at the same time (I've only really seen this happen with BuyVM/Ramnode).

    FIO is very disk intensive (and occasionally it hangs on some configurations), we're working on a way to exclude certain tests from the run if a certain thing gives you any trouble.

    If any hosts want to test how the script might impact others on the node shoot me a PM so we have some real world data rather than making assumptions.

    Thanked by 1acarrascope
  • Nick_ANick_A Member, Top Host, Host Rep

    @serverbear said: Multiple people are running it on the same node at the same time (I've only really seen this happen with BuyVM/Ramnode).

    Yeahhhh...

  • NanoG6NanoG6 Member
    edited September 2012

    To be honest I don't really understand what serverbear FIO benchmark is, yet it is the longest part. @serverbear can you add an option to skip that one?

    --edit

    @serverbear said: we're working on a way to exclude certain tests from the run if a certain thing gives you any trouble.

  • serverbearserverbear Member
    edited September 2012

    @NanoG6 said: To be honest I don't really understand what serverbear FIO benchmark is, yet it is the longest part. @serverbear can you add an option to skip that one?

    At the moment it sets the number of processes based on the cores your CPU has, we're doing a bit of testing to determine whether we can get the same sort of results/comparison by just running 1 process (which will speed it up / reduce load).

    For example 1 core / process takes ~7mins, 4 cores / processes takes ~ 30mins on a Linode.

  • Well, for me, I would want to apologize for every host that ever benchmarked by me and causing high io or trouble :)

  • Okay, we should have fixed any issues with the FIO test taking longer than normal. Not sure if you want to risk trying again Chan :)

  • and that, ladies and gentleman is the reason why i don't run serverbear benchmark on my vps

    @serverbear said: For example 1 core / process takes ~7mins, 4 cores / processes takes ~ 30mins on a Linode.

    7 mins that's more like it. I can't wait for this. =-)

  • The problem is that Serverbear on Linux-VServer uses 100% on 24 (all available) cores and all available IO across the full RAID array - not a good idea to run on it.

    Thanked by 1Oliver
  • KuJoeKuJoe Member, Host Rep
    edited September 2012

    @fileMEDIA said: I think it´s not a good business if you suspend a vps when it using some io ressources in hour or so.

    So instead of all clients following the TOS, providers should make changes to few abusive VPSs to make them in compliance while they impact other users who are able to follow the rules? That doesn't seem like good business practices to me.

  • @William said: The problem is that Serverbear on Linux-VServer uses 100% on 24 (all available) cores and all available IO across the full RAID array - not a good idea to run on it.

    It won't do this now for the FIO tests, we've limited it to 1 process instead of 1 per core.

  • Hehe one of my edis VPS is down atm, maybe it is on your node ;)

  • MaouniqueMaounique Host Rep, Veteran

    @rds100 said: Exactly, VRS doesn't have the features needed for VPNing.

    Actually, it has ssh...
    M

  • @serverbear said: It won't do this now for the FIO tests, we've limited it to 1 process instead of 1 per core.

    Would that be less accurate in any way? or do the old benchmark and the new single core limited one get the same results?

  • @AsadHaider said: Would that be less accurate in any way? or do the old benchmark and the new single core limited one get the same results?

    With my testing today 1 process vs multiple processes doesn't make much difference to the end avg result.

Sign In or Register to comment.