cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nitin Mehta <Nitin.Me...@citrix.com>
Subject Re: Review Request 17747: CLOUDSTACK-6023:Non windows instances are created on XenServer with a vcpu-max above supported xenserver limits
Date Fri, 07 Feb 2014 21:16:06 GMT
Animesh - done. Please cherry pick to 4.3

Daan - for pushing this to master I will have to revert your fix in master.

On 07/02/14 11:48 AM, "Animesh Chaturvedi" <animesh.chaturvedi@citrix.com>
wrote:

>Can I have the patch applied in 4.3-forward?
>
>> -----Original Message-----
>> From: Nitin Mehta [mailto:Nitin.Mehta@citrix.com]
>> Sent: Friday, February 07, 2014 11:33 AM
>> To: dev@cloudstack.apache.org
>> Cc: Harikrishna Patnala; Anthony Xu
>> Subject: Re: Review Request 17747: CLOUDSTACK-6023:Non windows
>> instances are created on XenServer with a vcpu-max above supported
>> xenserver limits
>> 
>> Hey Daan,
>> I guess you are talking about your fix submitted in master [1].
>> You might want to see Hari's latest patch [2] as well. It solves the
>>following
>> two things which I guess your fix misses.
>> 
>> 1. It sets the vcpu max to a hardcoded value (say 16) only when dynamic
>> scaling is enabled. Do note that by default dynamic scaling is disabled.
>> This is an important fix bcz only those who need dynamic scaling get
>> impacted.
>> 2. It has made the vcpu max configurable at cluster level which makes it
>> flexible for the admin to customize it depending on the load in his/her
>> cluster.(I guess even you guys mentioned keeping it configurable) In
>>case
>> there are issues with keeping it at 16, there is a flexibility to
>>change it to a
>> lower value during runtime.
>> 
>> Let me know if you have any concerns.
>> 
>> 
>> [1]
>> 
>>https://git-wip-us.apache.org/repos/asf?p=cloudstack.git;a=blobdiff;f=plu
>>gi
>> ns/hypervisors/xen/src/com/cloud/hypervisor/xen/resource/CitrixResourceB
>> ase
>> .java;h=cf5c6d6c7623e682e6c5bd66d829351b2bf6ad49;hp=200a72ff219d5214
>> ba3ebfc
>> 2c198517e849e03a9;hb=0839fbc;hpb=b3829e54d6b7af426f797ffb9fa54b4cd2
>> abffc0
>> 
>> [2] https://reviews.apache.org/r/17747/
>> 
>> Thanks,
>> -Nitin
>> 
>> On 06/02/14 10:10 PM, "Daan Hoogland" <daan.hoogland@gmail.com>
>> wrote:
>> 
>> >Animesh,
>> >
>> >I put in a patch that makes it double the number of assigned vcpu or
>> >16 whichever is smaller. it is on 4.3-forward
>> >
>> >On Fri, Feb 7, 2014 at 6:28 AM, Animesh Chaturvedi
>> ><animesh.chaturvedi@citrix.com> wrote:
>> >>
>> >>
>> >>> -----Original Message-----
>> >>> From: Daan Hoogland [mailto:daan.hoogland@gmail.com]
>> >>> Sent: Thursday, February 06, 2014 6:33 AM
>> >>> To: Harikrishna Patnala
>> >>> Cc: Nitin Mehta; cloudstack
>> >>> Subject: Re: Review Request 17747: CLOUDSTACK-6023:Non windows
>> >>> instances are created on XenServer with a vcpu-max above supported
>> >>> xenserver limits
>> >>>
>> >>> we have hosts with 80 vms. 80*16 > 160 , which is spedcified in the
>> >>>xenserver  docs Joris came up with. That last part is not important
>> >>>to me but I am still  worried about the size of the statistics post
>> >>>by the members to the
>> >>>pool-
>> >>> master. If we can make sure we don't cross this boundary I am fine
>> >>>with not  making it optional. So to stress my point: even with a
>> >>>documented limit of
>> >>> 16 per vm there is also a limit of 160 per host. And the real limit
>> >>>is neither as  we can instantiate vms with 32 vcpu (even on
>> >>> 6.0.2 i think Joris?) the actual problem is in the internal xapi
>> >>>traffic.
>> >>>
>> >> [Animesh] xapi traffic issues need to be addressed by xen and outside
>> >>of cloudstack. From xen doc the vCPU  per host is 4000 not 160 (which
>> >>is logical processor/host). For 4.3 does it make sense to keep the
>> >>value lower to like 8 to reduce chances of overloading xapi. For 4.4
>> >>this can be reworked to a configurable or computed value bases on
>> >>different limits
>> >>>
>> >>>
>> >>> On Thu, Feb 6, 2014 at 1:47 PM, Harikrishna Patnala <
>> >>> harikrishna.patnala@citrix.com> wrote:
>> >>>
>> >>> >    This is an automatically generated e-mail. To reply, visit:
>> >>> > https://reviews.apache.org/r/17747/
>> >>> >
>> >>> > On February 6th, 2014, 8:41 a.m. UTC, *daan Hoogland* wrote:
>> >>> >
>> >>> > Harikrishna, I would like to see the intermediate option of scale
>> >>> > up
>> >>>to
>> >>> double the amount as well. Did you revert it? Is there a problem
>> >>>with this  approach?
>> >>> >
>> >>> > Otherwise your submission is fine, of course.
>> >>> >
>> >>> >  Hi Daan,
>> >>> > I did not revert any changes. We can put an option to scale upto
>> >>>double
>> >>> but why it is required if vm can go till 16 (if at all 16 is the
>> >>>correct limit).
>> >>> >
>> >>> >
>> >>> >
>> >>> > - Harikrishna
>> >>> >
>> >>> > On February 5th, 2014, 5:19 p.m. UTC, Harikrishna Patnala wrote:
>> >>> >   Review request for cloudstack and Nitin Mehta.
>> >>> > By Harikrishna Patnala.
>> >>> >
>> >>> > *Updated Feb. 5, 2014, 5:19 p.m.*
>> >>> >  *Bugs: *
>> >>> > CLOUDSTACK-
>> 6023<https://issues.apache.org/jira/browse/CLOUDSTACK-
>> >>> 6023>
>> >>> >  *Repository: * cloudstack-git
>> >>> > Description
>> >>> >
>> >>> > CLOUDSTACK-6023:Non windows instances are created on XenServer
>> >>> > with a vcpu-max above supported xenserver limits
>> >>> >
>> >>> > VCPUs-max value is changed to 16 and only when dynamic scaling
is
>> >>> enabled.
>> >>> >
>> >>> >   Diffs
>> >>> >
>> >>> >    -
>> >>>
>> >>>plugins/hypervisors/xen/src/com/cloud/hypervisor/xen/resource/CitrixR
>> >>>eso
>> >>> urceBase.java
>> >>> >    (bf9b068)
>> >>> >
>> >>> > View Diff <https://reviews.apache.org/r/17747/diff/>
>> >>> >
>> >>>
>> >>>
>> >>>
>> >>> --
>> >>> Daan
>> >
>> >
>> >
>> >--
>> >Daan
>


Mime
View raw message