Howdy, Stranger!

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


Testing a Contabo VPS again after a while.... it's not THAT bad :D - Page 3
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.

Testing a Contabo VPS again after a while.... it's not THAT bad :D

13»

Comments

  • @arda said:

    @vitobotta said:

    @arda said:
    @vitobotta So far, so good. Only had issues once in both my VPSes in half a year, and it was known issue, and it got fixed in about a half a day, responsibe and active provider here as well. Happy customer here :blush:

    Awesome to hear. What about CPU steal? Since we were talking about that for Contabo

    Would this lines from top while I'm exporting my backup and compressing help?

    Exporting:

    %Cpu(s): 16.6 us, 73.8 sy, 0.0 ni, 9.2 id, 0.0 wa, 0.0 hi, 0.3 si, 0.0 st

    Compressing:

    %Cpu(s): 67.8 us, 1.4 sy, 0.0 ni, 30.7 id, 0.0 wa, 0.0 hi, 0.2 si, 0.0 st

    It rarely becomes 0.2 but almost always 0.0

    That's great news. Would you mind running the script I posted above for a day and see if you get alerts for steal higher than say 5%?

  • @vitobotta said:

    @arda said:

    @vitobotta said:

    @arda said:
    @vitobotta So far, so good. Only had issues once in both my VPSes in half a year, and it was known issue, and it got fixed in about a half a day, responsibe and active provider here as well. Happy customer here :blush:

    Awesome to hear. What about CPU steal? Since we were talking about that for Contabo

    Would this lines from top while I'm exporting my backup and compressing help?

    Exporting:

    %Cpu(s): 16.6 us, 73.8 sy, 0.0 ni, 9.2 id, 0.0 wa, 0.0 hi, 0.3 si, 0.0 st

    Compressing:

    %Cpu(s): 67.8 us, 1.4 sy, 0.0 ni, 30.7 id, 0.0 wa, 0.0 hi, 0.2 si, 0.0 st

    It rarely becomes 0.2 but almost always 0.0

    That's great news. Would you mind running the script I posted above for a day and see if you get alerts for steal higher than say 5%?

    Just started on tmux, because I'm a lazy person.

    bash cpu-steal-check.sh 5 30 [email protected]

    I believe is to check for steal, 5% for every 30 secs. Would it work?

  • @arda said:

    @vitobotta said:

    @arda said:

    @vitobotta said:

    @arda said:
    @vitobotta So far, so good. Only had issues once in both my VPSes in half a year, and it was known issue, and it got fixed in about a half a day, responsibe and active provider here as well. Happy customer here :blush:

    Awesome to hear. What about CPU steal? Since we were talking about that for Contabo

    Would this lines from top while I'm exporting my backup and compressing help?

    Exporting:

    %Cpu(s): 16.6 us, 73.8 sy, 0.0 ni, 9.2 id, 0.0 wa, 0.0 hi, 0.3 si, 0.0 st

    Compressing:

    %Cpu(s): 67.8 us, 1.4 sy, 0.0 ni, 30.7 id, 0.0 wa, 0.0 hi, 0.2 si, 0.0 st

    It rarely becomes 0.2 but almost always 0.0

    That's great news. Would you mind running the script I posted above for a day and see if you get alerts for steal higher than say 5%?

    Just started on tmux, because I'm a lazy person.

    bash cpu-steal-check.sh 5 30 [email protected]

    I believe is to check for steal, 5% for every 30 secs. Would it work?

    Yep, do you have a mailer like postfix set up?

  • @vitobotta said:

    @arda said:

    @vitobotta said:

    @arda said:

    @vitobotta said:

    @arda said:
    @vitobotta So far, so good. Only had issues once in both my VPSes in half a year, and it was known issue, and it got fixed in about a half a day, responsibe and active provider here as well. Happy customer here :blush:

    Awesome to hear. What about CPU steal? Since we were talking about that for Contabo

    Would this lines from top while I'm exporting my backup and compressing help?

    Exporting:

    %Cpu(s): 16.6 us, 73.8 sy, 0.0 ni, 9.2 id, 0.0 wa, 0.0 hi, 0.3 si, 0.0 st

    Compressing:

    %Cpu(s): 67.8 us, 1.4 sy, 0.0 ni, 30.7 id, 0.0 wa, 0.0 hi, 0.2 si, 0.0 st

    It rarely becomes 0.2 but almost always 0.0

    That's great news. Would you mind running the script I posted above for a day and see if you get alerts for steal higher than say 5%?

    Just started on tmux, because I'm a lazy person.

    bash cpu-steal-check.sh 5 30 [email protected]

    I believe is to check for steal, 5% for every 30 secs. Would it work?

    Yep, do you have a mailer like postfix set up?

    Nope, but stdout should suffice, right?

  • sthostingsthosting Member, Patron Provider

    @Arkas said:
    BTW, I would also like to add that another German provider that has served me exceptionally well and is not mentioned a lot here is @sthosting . They are worth looking into. Once again, as is always the case. I am not affiliated with them or any provider or service, these are all my opinions.

    Thank you so much for the feedback :)

    @Ympker Thanks to you too :)

    And @vitobotta Yep we can‘t hold the pricing with contabo, but our focus is on a very good customer support and no overprovisioned Hostsystems :) - but you are always welcome back to get our customer again :)

    Thanked by 1Ympker
  • @vitobotta said:

    @JabJab said:

    @vitobotta said: How can they be so cheap?

    You probably want to wait a little ~~
    https://lowendtalk.com/discussion/comment/3580277/#Comment_3580277

    February right? That's cool, I can wait. Do you have any servers with them already?

    I used RackNerd VPSes before, but they became OOS in EU for very long time so I needed alternative. I tried GreenCloud 3 months ago and so far I'm very happy. I have 2x VPS in NL and 1x VPS in DE.

    • There was problem networking in DE node week ago and server was painfully slow to work with for couple of hours
    • Other than that there wasn't any downtime
    • Performance is VERY solid for its price. I noticed way more consistent performance than on Racknerd VPS in AMS, but it all depends on node with LET providers so YMMV.
    • Support is VERY fast, although keep your expectation low as price is low.
    • I saw worse routing on one VPS in AMS compared to another one (Ping to Cloudflare 1ms vs 11ms) and asked if they could change to the same route, they didn't do it. 11ms isn't bad in any means, but its the same colo/DC (YISP) so kinda weird. With RackNerd AMS its always 1ms.
    • OS reinstalls are limited to 30. Likely its lifetime limit, as it didn't reset for me. I dunno what will happen if I will exhaust this limit, maybe it exists only to protect from abusers and you just need to send ticket? Maybe.
  • I actually have two VPS in Contabo and have no complaints at all. For more than a year, I can say that I am really happy with the performance and, overall, with the price as well. I am not sure why people do not like Contabo, so if someone can offer some light to me, please feel free.

    These are the ones I have.

    CLOUD VPS S
    €5.99 / month
    4 vCPU Cores
    8 GB RAM
    50 GB NVMe
    or 200 GB SSD
    1 Snapshot
    32 TB Traffic

    Model name: AMD EPYC 7282 16-Core Processor

    In this one, I'm hosting my own virtual desktop infrastructure, which I use almost every day. Very good performance for being servers with "no GPU."

    STORAGE VPS L
    €17.49 / month
    6 vCPU Cores
    16 GB RAM
    1.6 TB SSD
    600 Mbit/s Port
    32 TB Traffic

    Model name: AMD EPYC 7282 16-Core Processor

    This one is for business purposes, but I have to say that I am really surprised with the performance of the new Storage series. I had an old series storage server that I needed to cancel after a month about a year ago. The new series is equipped with the same processors as the normal Cloud series Contabo. Great performance!

  • LOL since I added the monitoring for the CPU steal for when it goes above 10% I am getting alerts all the time :joy: I contacted support to see what they say/do about it. The apps work decently well but anyway, nice if we can improve it somehow. I simplified my script to monitor the CPU steal by using the sar command:

    #!/bin/bash
    
    [ "${FLOCKER}" != "$0" ] && exec env FLOCKER="$0" flock -en "$0" "$0" "$@" || :
    
    THRESHOLD=10.0
    EMAIL=...
    
    while true; do
      sudo sar 1 1 -o > /tmp/steal
    
      CPU_Steal=$(cat /tmp/steal | grep Average | awk '{print $7}')
    
      echo "$(date) $CPU_Steal"
    
      if [ $(echo "$CPU_Steal > $THRESHOLD" | bc -l) == 1 ]; then
        echo "High CPU Steal \"$CPU_Steal%\" on $(hostname) as on $(date)" |
        mail -s "Alert: High CPU Steal $CPU_Steal%" $EMAIL
      fi
    done
    
    Thanked by 1ehab
  • well i dont even have to wait a month i believe

  • ehabehab Member
    edited January 2023

    @vitobotta said:

    what is the value of $FLOCKER var ?

  • nullroutenullroute Member, Host Rep

    Contabo is rubbish just stay away.

  • @ehab said:

    @vitobotta said:

    what is the value of $FLOCKER var ?

    It basically gets set to the current script name. It's just a trick I use on all the scripts when I want to make sure there is only an instance running at any time, like backups etc.

    Thanked by 2ehab arda
  • @true_nullroute said:
    Contabo is rubbish just stay away.

    I'll keep it for now until @greencoludvps has the new servers in Frankfurt, which should be in February. I host Mastodon and my blog on this server and so far it's been kinda ok.

  • Higher steal is average at Contabo. You can ask support to migrate you to a less loaded node but after a short time it will be loaded. They provisioned my new VPS M NVMe to an empty node, where I reached 10--20 GBps NVMe speeds, 950 single and 4900 multicore GB5 scores.

    After a day single core 600, multi 2300 and speeds 1--2 Gbps. Their VPS S seems to be unbeatable for their price but if someone wants better performance netcup is better with their RS line. The support very same at Contabo and netcup. If a problem fixable easily or a lot of users affected they are very quick but for non-general problems (IPv6 routing errors, unusable network segments in specific time-windows) they very very hard.

  • @adns said:
    Higher steal is average at Contabo. You can ask support to migrate you to a less loaded node but after a short time it will be loaded. They provisioned my new VPS M NVMe to an empty node, where I reached 10--20 GBps NVMe speeds, 950 single and 4900 multicore GB5 scores.

    After a day single core 600, multi 2300 and speeds 1--2 Gbps. Their VPS S seems to be unbeatable for their price but if someone wants better performance netcup is better with their RS line. The support very same at Contabo and netcup. If a problem fixable easily or a lot of users affected they are very quick but for non-general problems (IPv6 routing errors, unusable network segments in specific time-windows) they very very hard.

    Yeah it’s quite high quite often for me now. Support hasn’t replied yet.

  • jsgjsg Member, Resident Benchmarker
    edited January 2023

    @adns said:
    Higher steal is average at Contabo. You can ask support to migrate you to a less loaded node but after a short time it will be loaded. They provisioned my new VPS M NVMe to an empty node, where I reached 10--20 GBps NVMe speeds, 950 single and 4900 multicore GB5 scores.

    After a day single core 600, multi 2300 and speeds 1--2 Gbps. Their VPS S seems to be unbeatable for their price but if someone wants better performance netcup is better with their RS line. The support very same at Contabo and netcup. If a problem fixable easily or a lot of users affected they are very quick but for non-general problems (IPv6 routing errors, unusable network segments in specific time-windows) they very very hard.

    Not that I want to cover for Contabo - yes, they do have what I call a performance lottery due to, I strongly assume, lackluster monitoring and abuse detection, but ...

    For one I think it's also a perspective problem. People hear ZEN and expect high performance. The solution: don't see what you get as "Ryzen minus a whole lot" but rather as "E5 v3 or 4 plus quite some, and at a great price".
    Also keep in mind that Epyc is not the speed demon processor that Ryzen is; Epyc is basically all about getting lots of cores into a chip and at acceptable performance and energy consumption levels.

    I can remember when ZEN entered hosting. Back then most (affordable) VPS were E5 v2 or, at best, v4 and a typical hardware core count was 8 (times 2 because those servers usually had 2 sockets)- And BANG, with Epyc one suddenly could have 32 hardware cores at a reasonable price and decent power consumption. This basically boiled down to providers being able to quadruple the number of users per server and at lower cost/VPS.
    But of bloody course that's not the whole story. Quadrupling users per server also meant quadrupling disk and network access/IO/throughput ... but without network cards, routers, bandwidth and disks capable to quadruple easily (although NVMe became more and more available and used).

    Also: What is Contabo expected to do (other than to bloody get their monitoring up to par)?
    You see, there is a simple and in a way brutal mechanism behind it: If you put 1 user per hardware core that user gets excellent performance - but at a high price, if you put 1 user per hardware thread he still gets very nice performance (albeit significantly less that 1/2 core) - but still at a price that doesn't sell well.
    Now, enter the world of low-cost VPS hosting, where the equation looks more like x (quite a few, even 8 is not unheard of) users/VPS vCores per hw core - but for a really low price.

    I happened to have quite direct access to some managers at Contabo and (so it seemed) they had been told to be quite frank and open to me. So, for instance, I saw with my own eyes how many vCores per hardware core they have; and frankly, I was almost shocked how low that number was. But still, there are many hardware cores on a dual socket Epyc server ... so, unless you monitor your users tightly and cut down on (often not even intentional) abusers, you'll end up with quite some "steal" (to use what many seem to be concerned about although it tells way less than people believe).

    BUT "many (very decent) cores per server" also allows for attractive pricing!

    Turn it as you like but fact is that every provider has to decide for some spot on that axis for his products, which at the end of the day translates to "you can get a cheap VPS or you can get a fast VPS - but not both. I've gained quite some insight into providers operations and some even wanted (and usually took) my advice, and I can tell you that Contabo has made smart decisions and found a good sweet spot on the axis between cheap and performance. And I can also tell you that they use really good quality hardware.

    If you want zero "steal" and Ryzen performance, just get a Ryzen based VDS but be prepared to pay a price far above ca. €6/month.
    If you want "cheap, cheap, cheap!", look for an E5v2 based VPS, there are still quite a lot around and while those processors don't come anywhere near an Epyc, let alone a Ryzen they are not sh_tty and damn good enough for many jobs.
    If you want a decent compromise look, to name 2 examples, at Contabo and @terrahost, the former more focused on low price, the latter more focused on performance.

    And, maybe the most important factor: look for what you really need (as opposed to number porn).

    Thanked by 2maverick terrahost
  • Absolutely brilliant comment, @jsg !

    I also wanted to add, that it's not really important what the %steal is, 0%, 10%, 50%, AS LONG as you don't hit 100% CPU usage with your own workload. Because, that is the only thing that matters. When you deplete your available CPU resources, only then your service gets into trouble.

    I've already attached graphs from one of my VPS with the above average steal. What I can also tell is that I have stable workload there that used 10-15% CPU before that CPU steal has risen so much. So, now my graphs show that I use up to 20%, it just means that due to extra steal, the VPS is now slightly slower and thus utilized more. But, everything works just as well and stable as before.

    Thanked by 1jsg
  • @vitobotta I ran it almost for a day, for 30 secs interval. I even created 2 backup jobs on my GreenCloud BudgetKVMDE-2 server, on Debian.

    For each intervals, CPU steal values are either "0" or "0.016....", mostly 0.

  • I was told by GreenCloudVPS sales that the new CPUs in February will be only Premium VPS, so I got one of the budget ones now ($45/year is amazing). Thanks again @arda for suggesting them. Here's the YABS (4 cores, 8GB in Frankfurt)

    # ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
    #              Yet-Another-Bench-Script              #
    #                     v2022-12-29                    #
    # https://github.com/masonr/yet-another-bench-script #
    # ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## #
    
    Mon Jan  9 04:11:14 PM GMT 2023
    
    Basic System Information:
    ---------------------------------
    Uptime     : 0 days, 0 hours, 5 minutes
    Processor  : Intel(R) Xeon(R) CPU E5-2698 v4 @ 2.20GHz
    CPU cores  : 4 @ 2199.984 MHz
    AES-NI     : ✔ Enabled
    VM-x/AMD-V : ✔ Enabled
    RAM        : 7.8 GiB
    Swap       : 4.0 GiB
    Disk       : 55.1 GiB
    Distro     : Ubuntu 22.04.1 LTS
    Kernel     : 5.15.0-47-generic
    
    fio Disk Speed Tests (Mixed R/W 50/50):
    ---------------------------------
    Block Size | 4k            (IOPS) | 64k           (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 161.11 MB/s  (40.2k) | 639.64 MB/s   (9.9k)
    Write      | 161.54 MB/s  (40.3k) | 643.01 MB/s  (10.0k)
    Total      | 322.66 MB/s  (80.6k) | 1.28 GB/s    (20.0k)
               |                      |
    Block Size | 512k          (IOPS) | 1m            (IOPS)
      ------   | ---            ----  | ----           ----
    Read       | 713.79 MB/s   (1.3k) | 724.84 MB/s    (707)
    Write      | 751.72 MB/s   (1.4k) | 773.12 MB/s    (755)
    Total      | 1.46 GB/s     (2.8k) | 1.49 GB/s     (1.4k)
    
    iperf3 Network Speed Tests (IPv4):
    ---------------------------------
    Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping
    -----           | -----                     | ----            | ----            | ----
    Clouvider       | London, UK (10G)          | 1.22 Gbits/sec  | 472 Mbits/sec   | 13.8 ms
    Scaleway        | Paris, FR (10G)           | 2.68 Gbits/sec  | 2.80 Gbits/sec  | 16.1 ms
    Clouvider       | NYC, NY, US (10G)         | 1.28 Gbits/sec  | 714 Mbits/sec   | 88.4 ms
    
    iperf3 Network Speed Tests (IPv6):
    ---------------------------------
    Provider        | Location (Link)           | Send Speed      | Recv Speed      | Ping
    -----           | -----                     | ----            | ----            | ----
    Clouvider       | London, UK (10G)          | 7.75 Gbits/sec  | 470 Mbits/sec   | 13.8 ms
    Scaleway        | Paris, FR (10G)           | busy            | 2.98 Gbits/sec  | 16.4 ms
    Clouvider       | NYC, NY, US (10G)         | 1.63 Gbits/sec  | 2.15 Gbits/sec  | 87.3 ms
    
    Geekbench 5 Benchmark Test:
    ---------------------------------
    Test            | Value
                    |
    Single Core     | 719
    Multi Core      | 2668
    Full Test       | https://browser.geekbench.com/v5/cpu/19794562
    
    YABS completed in 6 min 18 sec
    
    Thanked by 2maverick arda
  • @arda said:
    @vitobotta I ran it almost for a day, for 30 secs interval. I even created 2 backup jobs on my GreenCloud BudgetKVMDE-2 server, on Debian.

    For each intervals, CPU steal values are either "0" or "0.016....", mostly 0.

    That’s awesome! Especially considering the prices

    Thanked by 1arda
  • Transferring 26GB from Contabo to GreenCloudVPS is taking ages due to the 200Mb/s port speed with Contabo. Looking at the YABS above you can see that it's much better with GC.

  • @arda what a difference... steal is basically constant at 0% with GreenCloud. Impressive for that price

    Thanked by 1arda
  • @vitobotta said:
    @arda what a difference... steal is basically constant at 0% with GreenCloud. Impressive for that price

    I'm glad you're happy with your new server 🙌

  • @arda said:

    @vitobotta said:
    @arda what a difference... steal is basically constant at 0% with GreenCloud. Impressive for that price

    I'm glad you're happy with your new server 🙌

    Let's see if it stays that way :D

    Thanked by 1arda
  • Btw does greencloud have snapshot feature? How many snapshot we can keep there?

  • @akhfa said:
    Btw does greencloud have snapshot feature? How many snapshot we can keep there?

    Nope, from what I could see, at least at the budget KVM sale servers.

    Thanked by 1akhfa
  • @akhfa said:
    Btw does greencloud have snapshot feature? How many snapshot we can keep there?

    The control panel is super basic, I don't see that functionality either

    Thanked by 1akhfa
  • @vitobotta said:
    Got a simple script working nicely https://gist.github.com/vitobotta/363fc042d1e2791ceecd3cc081c44514

    I start it at boot with

    /path/to/cpu-steal <threshold in %> <delay between checks> <email address for notifications>
    

    Stupid simple but it works :D

    In your script, you'd be better to set currentline initially before the while loop, and then change line 14 to previousStats="$currentStats"

    That way, you're always using the exact previous values rather than possibly missing some.

    Also, honestly I'd recommend having a slower polling interval than 1 second. You're only going to get 100 samples maximum per second per CPU, and at most 1% accuracy. It'd be less prone to jitter if you checked once per minute.

  • @ralf said:

    @vitobotta said:
    Got a simple script working nicely https://gist.github.com/vitobotta/363fc042d1e2791ceecd3cc081c44514

    I start it at boot with

    /path/to/cpu-steal <threshold in %> <delay between checks> <email address for notifications>
    

    Stupid simple but it works :D

    In your script, you'd be better to set currentline initially before the while loop, and then change line 14 to previousStats="$currentStats"

    That way, you're always using the exact previous values rather than possibly missing some.

    Also, honestly I'd recommend having a slower polling interval than 1 second. You're only going to get 100 samples maximum per second per CPU, and at most 1% accuracy. It'd be less prone to jitter if you checked once per minute.

    See other comment above, I simplified the script a lot

Sign In or Register to comment.