mesos-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From haosdent <haosd...@gmail.com>
Subject Re: how to change mesos resources
Date Sun, 10 Apr 2016 13:27:07 GMT
Glad to hear you solve your problem and share your use case! You are always
welcome.
On Apr 10, 2016 7:48 PM, "Stefano Bianchi" <jazzista88@gmail.com> wrote:

> YYYYYEEEESSSSS haosdent!
> that exactly works! And is exactly what i mean!!! Thank you a lot, thank
> you all, and sorry if i asked you so many times.
>
> Maybe is better if i present myself to you.
>
> My name is Stefano Bianchi, Italian student of Telecommunication
> engineering, i'm currently doing my master thesis at INFN European Nuclear
> physics institute in Bologna city. I'm working on Tier 1 CNAF
> infrastructure where the cloud infrastructure is based on an Openstack IaaS.
> Here i have run 6 virtual machines 3 masters, for high availability, and 3
> slaves, on one slave i'm running mesos DNS.
> the final goal should be the implementation of a PaaS for science European
> community which allows the exploitation of distributed datacenter resources
> for some computations or experiments. For this purpose i have to integrate
> projcect calico plugin in on mesos, in roder to distribute the tasks among
> different network in a transparent way.
>
> I'm describing my master thesis since i saw that many of you are
> Professors and engineers, so you could have interest on that.
>
> Thanks a lot, probbly in the next few months i will come to write here,
> for the moment i really appreciate your help.
> Thanks again!
>
>
> 2016-04-10 5:12 GMT+02:00 haosdent <haosdent@gmail.com>:
>
>> Good to see you solve the chronos problem. For your current problem, you
>> could edit the /etc/default/mesos-slave file and append the configuration
>> to it
>>
>> ```
>> MESOS_MASTER=131.154.96.32:5050
>> MESOS_HOSTNAME=131.154.96.52
>> MESOS_RESOURCES='cpu(*):1;mem(*):1300;disk(*):9000'
>> MESOS_LOG_DIR=/var/log/mesos
>> ```
>>
>> Then service mesos-slave restart
>>
>> On Sun, Apr 10, 2016 at 8:32 AM, Stefano Bianchi <jazzista88@gmail.com>
>> wrote:
>>
>>> Ops,i did not hide the IP addresses, sorry please don't hack :P
>>>
>>> 2016-04-10 2:30 GMT+02:00 Stefano Bianchi <jazzista88@gmail.com>:
>>>
>>>> i solved the chronos problem creating a hostname file in /etc/chronos
>>>> specifying the floating IP inside it.
>>>> Now all is working with changed resources, the only one problem that
>>>> persist is that after launching the command:
>>>>
>>>> mesos-slave --master=131.154.96.32:5050 --hostname=131.154.96.52
>>>> --resources='cpu(*):1;mem(*):1300;disk(*):9000' --log_dir=/var/log/mesos
>>>>
>>>> i still have some messages in command line and i cannot use the command
>>>> line.
>>>>
>>>> You're right haosdent i want run mesos-slave as service, but to do that
>>>> i shoulf press ctrl+c to be able to use the command line, then i commonly
>>>> use: service mesos-slave start command, but running it i obtain that the
>>>> slave registers to the master allocating the default resources.
>>>> How can i fix the changes of resources and run mesos-slave with service
>>>> command?
>>>>
>>>> Thanks again!
>>>>
>>>> 2016-04-09 22:14 GMT+02:00 Stefano Bianchi <jazzista88@gmail.com>:
>>>>
>>>>> Im sorry, the command was just:
>>>>>
>>>>> mesos-slave --master=131.154.xxx.xxx:5050 --hostname=131.154.xxx.xxx
>>>>> --resources='cpu(*):1;mem(*):1000;disk(*):8000'
>>>>>
>>>>> Where i specify the hostname with the floating IP.
>>>>>
>>>>> Without
>>>>> rm -f /tmp/mesos/meta/slaves/latest
>>>>> Actually i dont know what this command means...even if i have run it
>>>>> before.
>>>>> Il 09/apr/2016 20:51, "haosdent" <haosdent@gmail.com> ha scritto:
>>>>>
>>>>>> Yes, remove /tmp/mesos/meta/slaves/latest could cause this.
>>>>>>
>>>>>> On Sun, Apr 10, 2016 at 2:47 AM, Stefano Bianchi <
>>>>>> jazzista88@gmail.com> wrote:
>>>>>>
>>>>>>> Sorry maybe i was not clear.
>>>>>>> After that error with chronos i stopped the mesos slave and
>>>>>>> restarted it witj service mesos-slave start, and in this way
chronos job is
>>>>>>> working.
>>>>>>> Of i stop the slavr and start it again with the command that
changes
>>>>>>> the resources i get the error.
>>>>>>> In both cases i see in framework tab of mesos, that the hostname
>>>>>>> related to chronos is non resolvable "mesos2.novalocal"
>>>>>>> Im asking to my self if this can cause problem with this comamand:
>>>>>>>
>>>>>>> rm -f /tmp/mesos/meta/slaves/latest
>>>>>>>
>>>>>>>
>>>>>>> mesos-slave --master=131.154.xxx.xxx:5050 --hostname=131.154.xxx.xxx
>>>>>>> --resources='cpu(*):1;mem(*):1000;disk(*):8000'
>>>>>>> Il 09/apr/2016 19:28, "haosdent" <haosdent@gmail.com> ha
scritto:
>>>>>>>
>>>>>>> >No it does not happen normally, that message comes out only
after
>>>>>>> i launched the command to change the reaources. Maybe i should
create, as i
>>>>>>> made both for mesos and marathon, the hostname file for Chronos
in /etc/chronos
>>>>>>> and put it inside the public IP resolvable?
>>>>>>>
>>>>>>> Sorry could not get your points here. What happens if you create
a
>>>>>>> new chronos job and launch it? Is it still error?
>>>>>>>
>>>>>>> On Sun, Apr 10, 2016 at 1:03 AM, Stefano Bianchi <
>>>>>>> jazzista88@gmail.com> wrote:
>>>>>>>
>>>>>>>>
>>>>>>>> "I think maybe you clear your work_dir in agent. So those
messages
>>>>>>>> disappear now. Does this always happen when you create a
new chronos job?"
>>>>>>>>
>>>>>>>> No it does not happen normally, that message comes out only
after i
>>>>>>>> launched the command to change the reaources. Maybe i should
create, as i
>>>>>>>> made both for mesos and marathon, the hostname file for Chronos
in
>>>>>>>> /etc/chronos and put it inside the public IP resolvable ?
>>>>>>>>
>>>>>>>> Il 09/apr/2016 18:29, "haosdent" <haosdent@gmail.com>
ha scritto:
>>>>>>>> >
>>>>>>>> > >Framework with ID 'c01b5040-6079-4f55-8ef2-089f0d4d1987-0000'
>>>>>>>> does not exist on slave with ID 'ad490064-1a6e-415c-8536-daef0d8e3572-S20'.
>>>>>>>> >
>>>>>>>> > I think maybe you clear your work_dir in agent. So those
messages
>>>>>>>> disappear now. Does this always happen when you create a
new chronos job?
>>>>>>>> >
>>>>>>>> > > However i tried to launch your command but i still
unable to
>>>>>>>> use the terminal after that :P, this is another and minor
issue.
>>>>>>>> >
>>>>>>>> > I suppose you want to make mesos agent become a systemclt
>>>>>>>> service. If you use `apt/yum` install mesos from mesosphere,
it should OK
>>>>>>>> when you use  `systemctl` to start mesos agent. If you install
it from
>>>>>>>> source code, I think you need take some works to write systemctl
service
>>>>>>>> files for mesos.
>>>>>>>> >
>>>>>>>> > On Sun, Apr 10, 2016 at 12:21 AM, Stefano Bianchi <
>>>>>>>> jazzista88@gmail.com> wrote:
>>>>>>>>
>>>>>>>> >>
>>>>>>>>
>>>>>>>> >> i checked in framework, and i see that in correspondence
of
>>>>>>>> chronos host tab there is a non resolvable hostname... Why?
>>>>>>>> >>
>>>>>>>> >> 2016-04-09 18:14 GMT+02:00 Stefano Bianchi <jazzista88@gmail.com
>>>>>>>> >:
>>>>>>>>
>>>>>>>> >>>
>>>>>>>>
>>>>>>>> >>> Thanks haosdent.
>>>>>>>> >>> However i tried to see launch that command and
what i see is:
>>>>>>>> >>> 1) Resources are increased, this is what i expected,
good;
>>>>>>>> >>> 2) A Chronos Job that run each 2 minutes, is
allocated and then
>>>>>>>> "Staged";
>>>>>>>> >>> is i click on sandbox i obtain this message:
>>>>>>>> >>>
>>>>>>>> >>> Framework with ID 'c01b5040-6079-4f55-8ef2-089f0d4d1987-0000'
>>>>>>>> does not exist on slave with ID 'ad490064-1a6e-415c-8536-daef0d8e3572-S20'.
>>>>>>>> >>>
>>>>>>>> >>> What does it means? How can i solve it?
>>>>>>>> >>> However i tried to launch your command but i
still unable to
>>>>>>>> use the terminal after that :P, this is another and minor
issue.
>>>>>>>> >>>
>>>>>>>> >>>
>>>>>>>> >>> 2016-04-09 17:34 GMT+02:00 haosdent <haosdent@gmail.com>:
>>>>>>>>
>>>>>>>> >>>>
>>>>>>>>
>>>>>>>> >>>> Hi, @Stefano You could specify the flag
--log_dir when you
>>>>>>>> start Mesos Agent. For example:
>>>>>>>> >>>>
>>>>>>>> >>>> mesos-slave --master=MASTER_ADDRESS:5050
>>>>>>>> --hostname=slave_public_IP_i_set
>>>>>>>> --resources='cpu(*):1;mem(*):1000;disk(*):8000' --log_dir=/var/log/mesos
>>>>>>>> >>>>
>>>>>>>> >>>> So it would log to files under /var/log/mesos
instead of
>>>>>>>> console.
>>>>>>>> >>>>
>>>>>>>> >>>> On Sat, Apr 9, 2016 at 10:08 PM, Stefano
Bianchi <
>>>>>>>> jazzista88@gmail.com> wrote:
>>>>>>>>
>>>>>>>> >>>>>
>>>>>>>>
>>>>>>>> >>>>> Hi
>>>>>>>> >>>>> All is working fine!
>>>>>>>> >>>>> I can see more disk space and more RAM!
>>>>>>>> >>>>> The only thing that is different is
that with the Arjun last
>>>>>>>> command, i obtain a verbose output, i mean a lot of messages
continuously
>>>>>>>> shown, on command line. i would like to hide them and continue
to use the
>>>>>>>> command line, is it possible?
>>>>>>>> >>>>>
>>>>>>>> >>>>> Thanks again!
>>>>>>>> >>>>
>>>>>>>> >>>>
>>>>>>>> >>>>
>>>>>>>> >>>>
>>>>>>>> >>>> --
>>>>>>>> >>>> Best Regards,
>>>>>>>> >>>> Haosdent Huang
>>>>>>>> >>>
>>>>>>>> >>>
>>>>>>>> >>
>>>>>>>> >
>>>>>>>> >
>>>>>>>> >
>>>>>>>> > --
>>>>>>>> > Best Regards,
>>>>>>>> > Haosdent Huang
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> Best Regards,
>>>>>>> Haosdent Huang
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Best Regards,
>>>>>> Haosdent Huang
>>>>>>
>>>>>
>>>>
>>>
>>
>>
>> --
>> Best Regards,
>> Haosdent Huang
>>
>
>

Mime
View raw message