[ClusterLabs] Multiple apache services running

Ken Gaillot kgaillot at redhat.com
Mon Aug 3 17:29:40 EDT 2015


On 08/03/2015 12:08 PM, Vijay Partha wrote:
> Hi
> 
> For this question: Is it possible to control even for multiple instances of
> apache? like i have node 1 running 2 instances of apache and another node
> running 1 instance in apache. If my apache server goes down will all these
> instances be restarted?

What do you mean by instances? If you're using a single configuration
directory such as /etc/httpd, then chances are it's a single instance
from pacemaker's point of view, regardless of how many processes apache
spawns or how many websites it serves.

> Could you give me an example on how this can work. I mean commands that
> will make the above case pass.
> 
> Thanking You.
> 
> On Wed, Jul 29, 2015 at 10:34 AM, Vijay Partha <vijaysarathy94 at gmail.com>
> wrote:
> 
>> fine will have a look at it. thank you so much.
>> On 29 Jul 2015 22:56, "Ken Gaillot" <kgaillot at redhat.com> wrote:
>>
>>> On 07/29/2015 09:24 AM, Vijay Partha wrote:
>>>> could you give an idea on how active active could be achieved?
>>>
>>> Clusters From Scratch gives a simple walk-through of active/active. See
>>>
>>> http://clusterlabs.org/doc/en-US/Pacemaker/1.1-pcs/html-single/Clusters_from_Scratch/index.html
>>>
>>> The basic components are:
>>>
>>> 1. Fencing
>>>
>>> 2. Shared storage (DRBD in the walk-through, or could be a SAN or NAS,
>>> etc.)
>>>
>>> 3. Clustered file system (GFS2 in the walk-through) and optionally CLVM
>>> (if you want to use logical volumes with shared storage); these often
>>> require DLM (distributed lock manager) as well
>>>
>>> 4. Your service (apache in this case)
>>>
>>> 5. Some type of load balancer (multicast Ethernet in the walk-through,
>>> although something like haproxy is more often used in production)
>>>
>>>> On 29 Jul 2015 19:52, "Ken Gaillot" <kgaillot at redhat.com> wrote:
>>>>
>>>>> On 07/29/2015 09:03 AM, Vijay Partha wrote:
>>>>>> Thanks a lot Ken. for the second case whatever you said , is it
>>> possible
>>>>> to
>>>>>> control even for multiple instances of apache? like i have node 1
>>>>> running 2
>>>>>> instances of apache and another node running 1 instance in apache. If
>>> my
>>>>>> apache server goes down will all these instances be restarted?
>>>>>
>>>>> It depends on how you have those instances linked together.
>>>>>
>>>>> If they're independent -- one node runs www.A.com and another node
>>> runs
>>>>> api.A.com -- then there's no coordination necessary. Each can be
>>>>> monitored, stopped and started independently of the other.
>>>>>
>>>>> If your goal is to do active-passive failover (if the node serving
>>>>> www.A.com goes down, apache is started on another node), then
>>> pacemaker
>>>>> will handle that automatically.
>>>>>
>>>>> If your goal is to do active-active load-balancing (every node can
>>> serve
>>>>> requests to www.A.com simultaneously), that's a more complicated
>>>>> configuration, but pacemaker can do it.
>>>>>
>>>>> In any of the cases, it's generally not necessary to restart apache on
>>>>> all nodes. If it's not working on one node, it can be restarted there.
>>>>> You can manually disable and reenable the service if you want to force
>>> a
>>>>> restart everywhere.
>>>>>
>>>>>> On Wed, Jul 29, 2015 at 7:26 PM, Ken Gaillot <kgaillot at redhat.com>
>>>>> wrote:
>>>>>>
>>>>>>> On 07/29/2015 08:05 AM, Vijay Partha wrote:
>>>>>>>> Hi.
>>>>>>>>
>>>>>>>> Victor, Whatever you said i had tried and got it working. Thank you.
>>>>>>>>
>>>>>>>> Could you guys answer the following questions please?
>>>>>>>>
>>>>>>>> 1.) I have 2 apache services running on the same node. If the
>>> services
>>>>> go
>>>>>>>> down can pacemaker restart it?
>>>>>>>
>>>>>>> Yes, that's what pacemaker is for :)
>>>>>>>
>>>>>>> If by two services you mean two server instances of apache, then you
>>>>>>> would configure two apache resources in pacemaker. If you mean two
>>> web
>>>>>>> apps running in one apache server, you'd configure one apache
>>> resource
>>>>>>> in pacemaker.
>>>>>>>
>>>>>>> You would also need a monitor operation, which tells pacemaker to
>>>>>>> periodically check the health of the service.
>>>>>>>
>>>>>>> The migration-threshold and on-fail options determine what the
>>> cluster
>>>>>>> does on failure. See the documentation for possible values, but the
>>>>>>> default behavior is to try to restart the service on the same node.
>>>>>>>
>>>>>>> The documentation for all the options is at
>>>>>>>
>>>>>>>
>>>>>
>>> http://clusterlabs.org/doc/en-US/Pacemaker/1.1-pcs/html-single/Pacemaker_Explained/index.html
>>>>>>>
>>>>>>>> 2.) I have 2 apache services running on different nodes. Can i
>>> control
>>>>>>> both
>>>>>>>> the resources from a single node by making use of pacemaker?
>>>>>>>
>>>>>>> The various cluster tools work from any node, to affect any or all
>>>>>>> nodes. Most commonly you'd use one of the high-level cluster shells
>>>>>>> (crmsh or pcs). They have commands to enable/disable a resource,
>>> move a
>>>>>>> resource to a different node, put a node in standby mode, etc.





More information about the Users mailing list