cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Angus <paul.an...@shapeblue.com>
Subject RE: ACS 4.9.2 Management Server and Galera Cluster in CentOS 7
Date Wed, 19 Jul 2017 17:07:37 GMT
I did some work with Several Nines a couple of years ago to get Galera clusters certified by
Citrix for use with CloudPlatform.

They have an application call Cluster Control (https://severalnines.com/product/clustercontrol)
which simplifies the creation and running of Galera clusters (among other things).  It's been
quite a while since I looked at it.  Maybe of interest to some people.

I'm interested in the Java Connector HA - have you had many failovers and fail-backs which
it has had to deal with?




Kind regards,

Paul Angus

paul.angus@shapeblue.com 
www.shapeblue.com
53 Chandos Place, Covent Garden, London  WC2N 4HSUK
@shapeblue
  
 


-----Original Message-----
From: Simon Weller [mailto:sweller@ena.com.INVALID] 
Sent: 19 July 2017 14:37
To: users@cloudstack.apache.org
Subject: Re: ACS 4.9.2 Management Server and Galera Cluster in CentOS 7

Ivan,


We spend a fair bit of time a last year taking another look at running a Galera cluster for
our DB nodes and after a lot of testing,we decided it was too complicated to be reliable in
our case. We then moved on to testing the MySQL java connector HA support and we found it
to be broken, so we submitted a PR for that and it has been fixed since 4.9. We are using
the Java Connector HA and have been for about a year now in production and in our case, it
has worked well.


- Si


________________________________
From: Ivan Kudryavtsev <kudryavtsev_ia@bw-sw.com>
Sent: Tuesday, July 18, 2017 10:39 PM
To: users@cloudstack.apache.org
Subject: Re: ACS 4.9.2 Management Server and Galera Cluster in CentOS 7

Hello, Paul.

Yes, I see and realize the aftermath and actually I just haven't realized the thing about
get_lock, because main documentation (
http://docs.cloudstack.apache.org/projects/cloudstack-installation/en/4.9/management-server/)
Management Server Installation - Apache CloudStack ...<http://docs.cloudstack.apache.org/projects/cloudstack-installation/en/4.9/management-server/>
docs.cloudstack.apache.org
Overview¶ This section describes installing the Management Server. There are two slightly
different installation flows, depending on how many Management Server nodes ...



didn't mention that ACS requirement, and as I said before I was asked by a person to try deploy
the topology.

Just wondering if someone have thought about using Apache Zookeeper for distributed locks
coordination? We are using it with Apache Curator library in almost all our distributed projects
and there are recipes for the DLM:
http://curator.apache.org/curator-recipes/shared-reentrant-lock.html or Apache Curator Recipes<http://curator.apache.org/curator-recipes/shared-reentrant-lock.html>
curator.apache.org
public void acquire() Acquire the mutex - blocking until it's available. Note: the same thread
can call acquire re-entrantly. Each call to acquire must be balanced by ...



http://curator.apache.org/curator-recipes/shared-lock.html . I briefly took Apache Curator
Recipes<http://curator.apache.org/curator-recipes/shared-lock.html>
curator.apache.org
Fully distributed locks that are globally synchronous, meaning at any snapshot in time no
two clients think they hold the same lock. Note: unlike InterProcessMutex ...



a look at locking code and it seems Zk fits (if there are no other implications).

Also, as for single node configuration ZK server can be launched embedded in ACS (as it's
also a Java and widely used embedded for integrational testing and other embedded stuff) to
avoid additional configuration dependencies. Global configuration parameters might allow switching
from 127.0.0.1:28010/cloudstack (fake embedded) to a real cluster zk:// 1.1.1.1:2821,1.1.1.2:2821,1.1.1.3:2821/cloudstack.

Or, actually, db.ha.enabled, db.cloud.slaves options solve the problem and ZK is "overengineering"
approach?

Have a good day.

2017-07-19 1:14 GMT+07:00 Paul Angus <paul.angus@shapeblue.com>:

> Hi Ivan,
>
> Load balancer is a generic term, it doesn't have to distribute load 
> evenly.  In the case of a Galera cluster, it would have a primary 
> galera node to send traffic to, the other nodes would be fail over options.
>
> Every production environment should have a failover option for the 
> MySQL database in case of a physical hardware failure. Also, OSes can 
> become corrupted, and shared storage can become unavailable.
>
>
>
> Kind regards,
>
> Paul Angus
>
> paul.angus@shapeblue.com
> www.shapeblue.com<http://www.shapeblue.com>
[http://www.shapeblue.com/wp-content/uploads/2017/06/logo.png]<http://www.shapeblue.com/>

Shapeblue - The CloudStack Company<http://www.shapeblue.com/>
www.shapeblue.com
Rapid deployment framework for Apache CloudStack IaaS Clouds. CSForge is a framework developed
by ShapeBlue to deliver the rapid deployment of a standardised ...



> 53 Chandos Place, Covent Garden, London  WC2N 4HSUK @shapeblue
>
>
>
>
> -----Original Message-----
> From: Rafael Weingärtner [mailto:rafaelweingartner@gmail.com]
> Sent: 18 July 2017 18:17
> To: users@cloudstack.apache.org
> Subject: Re: ACS 4.9.2 Management Server and Galera Cluster in CentOS 7
>
> By failover as requirement, do you mean master-master replication and
> transparent DB access for ACS?
>
> On Tue, Jul 18, 2017 at 1:14 PM, Ivan Kudryavtsev <
> kudryavtsev_ia@bw-sw.com>
> wrote:
>
> > Hi, Paul. Thanks for advice. It seems that load balance is not an
> > option and the failover is the only case.
> > By the way does anyone tried, thought about implementing the case with
> > Zookeeper functionality?
> >
> > As I said, I personally haven't met the situation when load balancing
> > required or even failover required, but it might be for intensive
> > API-centric deployments it's an option?
> >
> > 2017-07-18 22:32 GMT+07:00 Paul Angus <paul.angus@shapeblue.com>:
> >
> > > Be careful with your load-balancing Ivan - Since CloudStack relies
> > > on MySQL's GET_LOCK and RELEASE LOCK, multi-master isn't supported.
> > >
> > > http://www.shapeblue.com/how-to-deploy-high-availability-
> > > cloudstack-with-mariadb-galera-cluster/
> > >
> > >
> > >
> > > Kind regards,
> > >
> > > Paul Angus
> > >
> > > paul.angus@shapeblue.com
> > > www.shapeblue.com<http://www.shapeblue.com>
> > > 53 Chandos Place, Covent Garden, London  WC2N 4HSUK @shapeblue
> > >
> > >
> > >
> > >
> > > -----Original Message-----
> > > From: Ivan Kudryavtsev [mailto:kudryavtsev_ia@bw-sw.com]
> > > Sent: 18 July 2017 16:08
> > > To: users@cloudstack.apache.org
> > > Subject: ACS 4.9.2 Management Server and Galera Cluster in CentOS 7
> > >
> > > Hello, community.
> > >
> > > A person asked me whether i have experience of deploying ACS 4.9 with
> > > Galera HA cluster and after I answered - "I don't" I decided to try
> > because
> > > I was unable to find relevant new article I wrote it as well. It
> explains
> > > how to deploy ACS 4.9.2 to Galera in CentOS 7. I published it under
> > Apache
> > > License, so If someone includes it in docs it would be great, please
> also
> > > don't hesitate notifying me with any bugs found and considerations.
> Hope
> > > it'll help to someone.
> > >
> > > Link:
> > > https://bitworks.software/blog/en/2017-07-18-mariadb-
> > > galera-cloudstack-ha-cluster.html
> > >
> > > Have a good day.
> > >
> > > --
> > > With best regards, Ivan Kudryavtsev
> > > Bitworks Software, Ltd.
> > > Cell: +7-923-414-1515
> > > WWW: http://bitworks.software/ <http://bw-sw.com/>
> > >
> >
> >
> >
> > --
> > With best regards, Ivan Kudryavtsev
> > Bitworks Software, Ltd.
> > Cell: +7-923-414-1515
> > WWW: http://bitworks.software/ <http://bw-sw.com/>
> >
>
>
>
> --
> Rafael Weingärtner
>



--
With best regards, Ivan Kudryavtsev
Bitworks Software, Ltd.
Cell: +7-923-414-1515
WWW: http://bitworks.software/ <http://bw-sw.com/>


Mime
View raw message