[Spacewalk-list] Large Spacewalk Deployments?

[Tomas Zajpt]

I was implement Spacewalk to manage 3-4k machines. You should use
proxy servers and osad cannot work for big numbers of machines. Random
access machines to proxies(rhn_check command) is absolutely necessary,
you should use rhnsd for times longer than hour or some other random
time access for less time.


Cheers,
Tomas

2013/11/14 Paul Robert Marino <prmarino1 at gmail.com>:
> Well one thing that can also  significantly help as well is liberal use of
> spacewalk proxies.
>
>
>
>
> -- Sent from my HP Pre3
>
> ________________________________
> On Nov 14, 2013 16:33, George <listmail.gg at gmail.com> wrote:
>
> From experience with only a couple of 100 hosts and 'old' versions of
> spacewalk (say 0.8 or 0.9) the spacewalk server gets hammered if you do
> for example and upgrade of 1 (or more) packages on a group of several
> 100 hosts at once ... (using osad so a "now" schedule) due to the fact
> it needs to receive the reporting that if the task failed or succeeded.
> My webinterface was unreachable for like 30 minutes and load on the box
> was 70 or so ...
> Note that at the time I was running it on dual core machine with 4G mem
> and oracle XE on the machine itself, so yes performance was not optimal.
>
> Now I have a setup with 8 spacewalk proxies in several global regions
> and around a 100 machines connected, but not really tried any major
> deployments like I mentioned above (soon to be adding up to around 1000
> hosts).
>
> I heard that for example puppet has some kind of backoff algorithm for
> this: if you get more requests than a certain max limit they will send
> the other requesters walking for a random time (also believe they have
> some kind of priority system, where a pushed update task gets priority
> over a random checkin from a server because the timer expired)
>
> Planning to combine spacewalk with puppet for certain tasks like this ...
>
> Regards,
>
> G.
>
> On 11/14/2013 07:30 PM, Paul Robert Marino wrote:
>> I'm fairly certain there are quite a few implementations of the
>> supported version RHN satellite on that scale, but there is a limit.
>> If I Remember correctly your numbers are well within the limit. There
>> are a few things you will have to keep in mind mainly the database
>> performance. I would suggest a minimum of a dual 16 core Opteron box and
>> possibly using an external database cluster with lots of small fast disks.
>>
>>
>> -- Sent from my HP Pre3
>>
>> ------------------------------------------------------------------------
>> On Nov 14, 2013 11:48, Andy Townsend <Andy_Townsend at symantec.com> wrote:
>>
>> Hey all,
>>
>> Does anyone know of or have any experience with using Spacewalk to
>> manage a large number of hosts? I’m looking at potentially using it to
>> push packages to some machines but we’re talking of around 5-10k nodes
>> here. Are there any spacewalk deployments out there managing that number
>> of hosts that anyone knows of or does anyone have any information on the
>> scalability aspects of Spacewalk.
>>
>> Cheers,
>>
>> Andy
>>
>>
>>
>> _______________________________________________
>> Spacewalk-list mailing list
>> Spacewalk-list at redhat.com
>> https://www.redhat.com/mailman/listinfo/spacewalk-list
>>
>
> _______________________________________________
> Spacewalk-list mailing list
> Spacewalk-list at redhat.com
> https://www.redhat.com/mailman/listinfo/spacewalk-list
>
> _______________________________________________
> Spacewalk-list mailing list
> Spacewalk-list at redhat.com
> https://www.redhat.com/mailman/listinfo/spacewalk-list



-- 

                    S pozdravem Tomas