Thanks, Ben. I didn't knew that OS balancer for shared sockets it not RR.

четверг, 6 июня 2013 г., 14:22:15 UTC+4 пользователь Ben Noordhuis написал:
>
> On Thu, Jun 6, 2013 at 7:43 AM, Vitaly Puzrin 
> <[email protected]<javascript:>> 
> wrote: 
> > In real life, we need sometime to update application on production 
> servers. 
> > A good 
> > practice is to do such things without service interruption. Usual 
> solution 
> > is binding 
> > new processes to the same address, and disabling old ones to accept new 
> > connections. 
> > In case of websockets, old processes can additionally signal clients to 
> > reconnect. 
> > 
> > Full algorythm is well described in nginx docs 
> > http://nginx.org/en/docs/control.html#upgrade 
> > 
> > Does it worth to create issue with request or such functionality is 
> specific 
> > for my needs? 
>
> I wouldn't reject such a pull request flat out but be prepared to go 
> through many rounds of review.  I've toyed with the idea of adding 
> rolling upgrade support and decided to save it for a rainy day... 
> there's a ton of edge cases you'll have to account for and more are 
> sure to pop up once the feature lands. 
>
> Secondly, I'm not sure how many people will actually use it.  I assume 
> most people have multi-machine setups where they simply take machines 
> out of the load balancer one at a time and upgrade at their leisure. 
>
> > The second question is about software RR scheduler, that i found in 
> cluster 
> > sources. Why 
> > it was was prefered over system's one? Will scheduler IPC add noticeable 
> > latency for server replies 
> > or not? 
>
> See [1] and the linked PR for rationale.  The commit log of [2] goes 
> into further detail. 
>
> In a nutshell, people observed very unevenly distributed loads on 
> Solaris and Linux systems.  Other operating systems are likely 
> affected as well. 
>
> Note that the RR scheduler is merely the default (except on Windows.) 
> The old approach, where the operating system distributes connections, 
> is still available and supported.  Grep the cluster API documentation 
> for 'cluster.schedulingPolicy' and 'NODE_CLUSTER_SCHED_POLICY'. 
>
> [1] https://github.com/joyent/node/issues/4435 
> [2] https://github.com/joyent/node/commit/e72cd41 
>

-- 
-- 
Job Board: http://jobs.nodejs.org/
Posting guidelines: 
https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines
You received this message because you are subscribed to the Google
Groups "nodejs" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/nodejs?hl=en?hl=en

--- 
You received this message because you are subscribed to the Google Groups 
"nodejs" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to