Announcement

Collapse
No announcement yet.

Clarify new CPU limits

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Clarify new CPU limits

    With the recent changes to CloudLinux and LVE manager, could you please clarify the new limits?

    I notice the documentation says that "speed" is relative to number of cores, so "100%" means 1 full core, "200%" is 2 full cores and so on. Does this now make the "nCPU" parameter redundant if "speed" is set?

    If "speed" is set to "200%" would "nCPU" be ignored?

    Finally, can you also please clarify that changes to these settings no longer requires a reboot, or is this still the case?

    Thanks in advance.

  • #2
    There has been no need for a reboot for quite some time now. About the new settings, while Im not 100% sure I guess that you still ned to specify 2 cores for 200% and 4 cores for 400%. It would however make sense to no longer need to specify the numer of cores as 200% is obviously 2 cores isn it ? If not then why would someone specify for instance 3 cores for 200% ?

    Comment


    • #3
      I appreciate the response, but with respect Id rather hear something conclusive instead of speculation on your part. Im a firm believer in not making any assumptions and theres certainly no harm in seeking verification if in doubt

      The reason I asked these questions was to get a definitive response rather than guesswork. The documentation has, over the years, been a tad confusing and sometimes lacking. An example of this is http://docs.cloudlinux.com/limits.html which still references CPU % instead of the new "speed" value. Another big example is the historical confusion over the number of CPUs vs the percentage of CPU power, with the former overriding the latter or taking precedence in certain situations. Not as intuitive as you would expect, but still something graspable. Since they have effectively replaced the previous "percentage of CPU power" variable with the new "speed" variable, it stands to reason that the nCPU *could* still maintain this same behaviour. Since this isn mentioned in the documentation or changelog I would just like to seek clarification on this. I agree, there is no reason to specify certain combinations of settings, but again - no harm in seeking clarification.

      In regards to the reboot being required, this setting was previously marked with an asterisk in the LVE manager stating that a reboot was required if changed, but now does not. This is, at least in my mind, no official confirmation that this is no longer required so I would also like confirmation of this. Please bear in mind not all users update their Kernels regularly and reboot to take advantage of newer changes.

      Can any CL staff please confirm the above? Thanks.

      Comment


      • #4
        No speculation, Im sure about not needing to reboot and am also sure about 200% needing 2 cores.

        I also asked if it couldnt be calculated automaticaly and if not why, so lets wait for CloudLinuxs answer now

        The new speed value just translates into % when you save its just a helper so you dont need to calculate the percentage manually.

        For the reboot it was necessary with the first versions only. I remember the asterisk but that must be at least a year ago, before cagefs became stable unless Im mistaken.

        Comment


        • #5
          And this is why I asked for CloudLinux to comment, to avoid situations like this

          I think we have our wires crossed somewhere and perhaps youve misinterpreted what Im asking. I wasnt asking if 200% "speed" needs 2 cores on the machine, which is how you seem to have interpreted it. I was asking if the nCPU setting is required if the speed setting is set. If nCPU is set, will it take precedence over "% of CPU" as it used to before they replaced this setting with "speed"?

          As simply as possible - If, for example, "speed" is set to 200% and "nCPU" is set to "2", which one (if any) takes priority? Of course, if you set "speed" to 200% and "nCPU" to "4", naturally both cant be valid at the same time so surely one has to take effect. This is what Im trying to establish here. CloudLinux will happily accept seemingly conflicting values, so this is something that needs to be clarified.

          With the previous limits, the documentation says the following:

          "The smallest of the two is used to define how much CPU power will be accessible to the customer."

          The new documentation says nothing of the sort.

          Clearer now?

          With regards to the reboot, this was visible in LVE manager not THAT long ago and definitely wasnt only a first few revision thing.

          Comment


          • #6
            I would love a clarification also. thanks.

            Comment


            • #7
              Sorry about this. I had understood that the only thing that had changed was the unit for 1%. However the documentation now says that ncpu is no longer required. Lets wait for cloudlinux to confirm about this. As if this is the case then why wasn the ncpu setting removed ?

              Comment


              • #8
                More info regarding speed is here: http://docs.cloudlinux.com/index.html?cpu_limits.html
                Basically speed lets you set CPU limit relative to a single core, or as a number of hrz.

                Comment


                • #9
                  Hello Igor,

                  According to the documentation the ncpu setting is no longer used. Can you please confirm that if I set 300% and 1 in ncpu it will use 3 CPUs ?

                  Comment


                  • #10
                    If you use CPU + NCPU --> it would work as before
                    If you use --speed --> NCPU ignored, not used.

                    Comment


                    • #11
                      In the lve cpanel interface 6 % and 2 cpus was converted to 197% and 2 cpus. Should
                      t the 2 cpus setting be empty ? Does the cpanel interface use speed and ignore the ncpu setting ? Or did it increase all my accounts settings ?

                      Comment


                      • #12
                        @Richard: "According to the documentation the ncpu setting is no longer used." unless were reading completely different documentation, you are wrong. It states nothing of the sort. Please stop making this statement because you will confuse other people as youve done multiple times in this thread already.

                        @Igor; "If you use CPU + NCPU --> it would work as before
                        If you use --speed --> NCPU ignored, not used."

                        Can you guys please, please, PLEASE just take some time to craft an actually comprehensible response and then update your documentation? With your above response, what are you actually saying? "If you use CPU + NCPU --> it would work as before." is what you are really saying here that if you havent upgraded, nothing will change? Isnt that already a little obvious?

                        "If you use --speed --> NCPU ignored, not used." - great. So the next logical question is, why does the nCPU option remain?

                        Can you please try and work with your clients here and provide some detailed documentation that makes sense of the situation? Im disappointed I had to come to the forum to seek clarification of something that shouldve been simple enough to explain in your documentation. Im also very disappointed that with all the posts we have made, you didnt address them properly and instead just gave short brief responses.

                        Comment


                        • #13
                          The documentation for the command line version is clear. LVE Manager documentation does need to be updated.

                          Although I did not understand Igors answer the first time I read it (missed the ignored part), it now makes sense, actually says everything and allows us to understand that the ncpu setting will likely be removed in a future release of LVE Manager as Igor said its currently ignored.

                          Thanks Igor, keep up the good work

                          Comment


                          • #14
                            @Richard: Are you purposefully being obtuse?? Ive mentioned LVE manager several times in this thread, and in fact my very first line in the very first post says the following:

                            > With the recent changes to CloudLinux and LVE manager....

                            To add to this, the "CPU limits" page of the documentation, which is the page you would expect to find detailed documentation on the implementation and settings regardless of command-line or UI interface, does not mention anything about nCPU being discontinued or no longer being used. Your comments about it being mentioned in the command line documentation are therefore irrelevant and unhelpful. For my own amusement I looked at the documentation for lvectl (which you claim confirms it is redundant and not being used) and it, also, does not mention anything about nCPU being redundant.

                            > Although I did not understand Igors answer the first time I read it (missed the ignored part), it now makes sense, actually says everything and allows us to understand that the ncpu setting will likely be removed in a future release of LVE Manager as Igor said its currently ignored.

                            You continue to make assumptions, false interpretations and attempts to "read between the lines" in the documentation and Igors responses. This may work for you and your business, but for me and ours I would like concrete and definitive responses here. I dont think thats unreasonable in the slightest. If CloudLinux are going to update their software, they need to update the documentation too. How they expect anyone to understand how their software works and to manage things properly without up to date and accurate documentation is beyond me.

                            I would really appreciate it if you would stop posting your assumptions and interpretations here. You are just going to confuse people looking for genuine and official information on the subject matter.

                            Comment


                            • #15
                              The new version was only released a couple of days ago, give them a bit of time and they will update the documentation

                              Comment

                              Working...
                              X