Howdy, Stranger!

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


CPU Steal time & SSD VPS suggestions required
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.

CPU Steal time & SSD VPS suggestions required

goodonegoodone Member
edited September 2018 in Help

Hi, I had several questions.

I am not experienced with Linux environment. So recently I have been having a problem with my VPS, it causing around 60% to 80% steal time under a simple page load before the page even starts loading. After passing of 15 to 20 sec, Reported page load only took place of 6 to 8sec, the page size is 2.01mb as per https://sitechecker.pro/page-size . Where I could only use CPU at the range of 10% to 20% percent per vcore. This WordPress site was doing perfect even in past months. Remember that when CPU steal time is high, it was not running the process for very long. Only at 5 sec, this happens. This is an SSD cached VPS and I had been with this provider for past 3 years+. They had been great so far. But, I think my server is dying.

So, Is it my VPS stalling or it's my site? If I use SSD VPS will this issue get fixed?

Next things, I wanted to know if there's any pure SSD VPS offer I can pick from a reliable provider. I have seen here some @ lowendbox, but not sure which one to go with. As I had pretty bad experience in this year with ZxHost Ceph cluster. (Didn't get the vps/refund).

I need at least 50gb to 60gb space.
4vcore, 4ram, 4/5tb bw, 1gbps (if possible), 3/4ips,
I will be using the vps for my staging sites purpose and will use cwp. (won't be high traffic).

Thanks

Comments

  • Have you opened a ticket? You might just be on a node with noisy neighbors, so your provider can probably fix that.

    st% in top implies that other VPS are hogging the CPU resource, effectively denying you access to it.

    Page load times could also just be an issue with disk I/O. There is more disk bandwidth available on full SSD (vs. SSD caching in front of spinners) so you might see better performance.

    Thanked by 1goodone
  • goodonegoodone Member
    edited September 2018

    @jiggawattz said:
    Have you opened a ticket? You might just be on a node with noisy neighbors, so your provider can probably fix that.

    st% in top implies that other VPS are hogging the CPU resource, effectively denying you access to it.

    Page load times could also just be an issue with disk I/O. There is more disk bandwidth available on full SSD (vs. SSD caching in front of spinners) so you might see better performance.

    Thanks for your response, You might be right about the denying access. Because that cms site is very simple and already well optimized.

    Is there any stress test I can do to check disk I/O? & Do you have any SSD openVZ/KVM vps provider to suggest?

    Yes, I have already notified support personally. Even though I had discussed with this issue previous month also, but I was being said it might be my site/panel, So I changed control panel to another and then it was same and again went back to the previous setup. Then later I found out about this st% thingy.

  • letboxletbox Member, Patron Provider

    @goodone said:

    @jiggawattz said:
    Have you opened a ticket? You might just be on a node with noisy neighbors, so your provider can probably fix that.

    st% in top implies that other VPS are hogging the CPU resource, effectively denying you access to it.

    you can use but the st may effect the performance.

    wget http://wget.racing/nench.sh

    Thanked by 1goodone
  • goodonegoodone Member
    edited September 2018

    @key900 said:

    @goodone said:

    @jiggawattz said:
    Have you opened a ticket? You might just be on a node with noisy neighbors, so your provider can probably fix that.

    st% in top implies that other VPS are hogging the CPU resource, effectively denying you access to it.

    you can use but the st may effect the performance.

    wget http://wget.racing/nench.sh

    Thanks, Just ran it. It was pretty slow while running the test, but here's the results. What can you tell me from these information?

    `[root@panel ~]# sh nench.sh

    nench.sh v2018.04.14 -- https://git.io/nench.sh

    benchmark timestamp: 2018-09-10 09:09:14 UTC

    Processor: Intel(R) Xeon(R) CPU L5520 @ 2.27GHz
    CPU cores: 4
    Frequency: 2261.050 MHz
    RAM: 2.9G
    Swap: -
    Kernel: Linux 2.6.32-754.3.5.el6.x86_64 x86_64

    Disks:
    xvda 80G SSD

    CPU: SHA256-hashing 500 MB
    23.751 seconds
    CPU: bzip2-compressing 500 MB
    29.918 seconds
    CPU: AES-encrypting 500 MB
    16.042 seconds

    ioping: seek rate
    min/avg/max/mdev = 116.7 us / 301.4 us / 240.6 ms / 3.74 ms
    ioping: sequential read speed
    generated 5.85 k requests in 5.00 s, 1.43 GiB, 1.17 k iops, 292.4 MiB/s

    dd: sequential write speed
    1st run: 92.03 MiB/s
    2nd run: 144.96 MiB/s
    3rd run: 164.03 MiB/s
    average: 133.67 MiB/s

    IPv4 speedtests
    your IPv4: xxx.xxx.xx.xxxx

    Cachefly CDN:         84.35 MiB/s
    Leaseweb (NL):        8.36 MiB/s
    Softlayer DAL (US):   61.83 MiB/s
    Online.net (FR):      4.29 MiB/s
    OVH BHS (CA):         10.19 MiB/s
    

    No IPv6 connectivity detected
    -------------------------------------------------`

  • letboxletbox Member, Patron Provider

    @goodone said:

    @key900 said:

    @goodone said:

    Well, Yes it slow.

  • JoeMeritJoeMerit Veteran
    edited September 2018

    the results you got for sha256 hashing, bzip2 compressing and AES encrypting are horrible.

    I currently have 4 cheap vpses and for sha256 hashing on them I always get below 5 seconds, for bzip2 around 8 seconds and for AES encrypting anywhere from 1.5 to 2.5 seconds.

    open a ticket and ask your provider to improve things, but have a backup ready and be prepared to cancel and move.

    Thanked by 2dfmcvn goodone
  • Yes, I have notified them and waiting for a proper response to help regarding this issue. :(
    It's just got horrible now: http://take.ms/wNKTo

  • @goodone - _f it is the provider I think it is They will fix it quickly.
    For future reference:

    Processor:    Intel(R) Xeon(R) CPU           L5520  @ 2.27GHz
    CPU cores:    2
    Frequency:    2261.050 MHz
    RAM:          1.8G
    Swap:         2.0G
    Kernel:       Linux 2.6.32-042stab133.2 x86_64
    
    Disks:
    xvda    100G  SSD
    
    CPU: SHA256-hashing 500 MB
        8.715 seconds
    CPU: bzip2-compressing 500 MB
        17.209 seconds
    CPU: AES-encrypting 500 MB
        7.355 seconds
    
    ioping: seek rate
        min/avg/max/mdev = 104.6 us / 186.8 us / 51.7 ms / 460.4 us
    ioping: sequential read speed
        generated 11.4 k requests in 5.00 s, 2.78 GiB, 2.28 k iops, 569.3 MiB/s
    
    dd: sequential write speed
        1st run:    393.87 MiB/s
        2nd run:    441.55 MiB/s
        3rd run:    411.99 MiB/s
        average:    415.80 MiB/s
    
    IPv4 speedtests
        your IPv4:    192.x.x.x
    
        Cachefly CDN:         70.41 MiB/s
        Leaseweb (NL):        14.68 MiB/s
        Softlayer DAL (US):   64.93 MiB/s
        Online.net (FR):      13.47 MiB/s
        OVH BHS (CA):         22.46 MiB/s
    
    IPv6 speedtests
        your IPv6:    2604:x:x:xxxx
    
        Leaseweb (NL):        15.84 MiB/s
        Softlayer DAL (US):   53.95 MiB/s
        Online.net (FR):      21.26 MiB/s
        OVH BHS (CA):         25.19 MiB/s
    -------------------------------------------------
    
  • WHO STOLE MY TIME ??? o(︶︿︶)o

  • goodone said: the page size is 2.01mb

    ^^^^^ there is your problem.

  • @willie said:

    goodone said: the page size is 2.01mb

    ^^^^^ there is your problem.

    This small size is not matters for cms site. Every site has cdn active with cache. Besides they always used to load within max 8sec Best load time 4 to 5sec before.

    Last night provider told me I am using storage VPS? so it could be the problem?. (wtf).
    Because I was given Xen HVM SSD Cached for web development tasks during purchase 3 years before. When I purchased. But past few months ago I have been moved without my awareness to a downgrade to different node version with the same capacity. And for this, I had been facing this issues past 5 months and I wasn't being told... Maybe, whenever you give total trust and vouch for a provider this is what happens.

    Still waiting on their inputs.

  • FrankZFrankZ Veteran
    edited September 2018

    goodone said: Last night provider told me I am using storage VPS?
    goodone said: Maybe, whenever you give total trust and vouch for a provider this is what happens.

    In a effort to be fair to the provider would it be safe to say that this was a slow HDD storage plan with the L5520 purchased circa 2015 that was temporarily upgraded to an X5670 and SSD array after a crash in 2016. The provider then moved these storage VPSes back to the L5520 and HDD array this past April?

    If so, it was pitched and sold by the provider as Raid5/50 storage with the L5520. The SSD and X5670 were nice, but in my understanding temporary. (As in nothing that good, at that price, is likely to be permanent)
    I miss mine too. :disappointed: but was glad to be part of the 42 23 this happened to.

    Good news is I think you can buy the SSD you know and love from the same provider for $7 a month plus ips.
    (when available :smiley: )

    Thanked by 1goodone
  • BharatBBharatB Member, Patron Provider

    @goodone said:
    I need at least 50gb to 60gb space.
    4vcore, 4ram, 4/5tb bw, 1gbps (if possible), 3/4ips,

    For $20 maybe we could do

    4 cores
    4GB RAM
    150GB SSD
    5TB bandwidth
    5 IPv4

  • goodone said: I need at least 50gb to 60gb space. 4vcore, 4ram, 4/5tb bw, 1gbps (if possible), 3/4ips,

    That is all doable except it's quite a high core to ram ratio, which means you'll be on shared cores. Typical too-slow Wordpress sites use badly configured plugins and you're better off optimizing that than throwing more hardware at it, and also the bloated 2.01MB download will inherently be slow.

    Anyway if you were at zxhost before and are ok with DE location and your traffic is mostly European, I'd try a Hetzner CX31 which has 8gb ram, 2 shared vcpu, and 80gb disk, at 9 euro/m. The CCX11 is similar but has dedicated vcpu at 20 euro/m. If you really need 4 vcpu go up to the 16GB vm, but that starts getting expensive enough that maybe you want to consider a dedicated server.

  • @FrankZ said:

    goodone said: Last night provider told me I am using storage VPS?
    goodone said: Maybe, whenever you give total trust and vouch for a provider this is what happens.

    In a effort to be fair to the provider would it be safe to say that this was a slow HDD storage plan with the L5520 purchased circa 2015 that was temporarily upgraded to an X5670 and SSD array after a crash in 2016. The provider then moved these storage VPSes back to the L5520 and HDD array this past April?

    If so, it was pitched and sold by the provider as Raid5/50 storage with the L5520. The SSD and X5670 were nice, but in my understanding temporary. (As in nothing that good, at that price, is likely to be permanent)
    I miss mine too. :disappointed: but was glad to be part of the 42 23 this happened to.

    Good news is I think you can buy the SSD you know and love from the same provider for $7 a month plus ips.
    (when available :smiley: )

    The provider has resolved the problem, They said they had to make some minor adjustment.

    I wasn't aware of any migration, except that crash. But thanks to them they have fixed it after putting it on their attention. Stealing time now much lower on CPU, and everything is back in functional again.

Sign In or Register to comment.