activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christopher Fogarty <Christopher.Foga...@Versiant.com>
Subject RE: Testing Master Slave on Shared File System
Date Fri, 29 Apr 2016 14:54:17 GMT
Quinn

Thanks. I could use NFS, but the master slave shared docs (very Spartan) indicated SAN. If
ext4 is not the proper filesystem. What should I use? I could use NFS, but I was concerned
over performance. Also, NFS in a VM environment would require a third server?

Node 1
Node 2
NFS server (I would not think it would be prudent to use nfs from one of the two nodes?)

Chris Fogarty

VP, System Engineering
Versiant Corporation
3700 Arco Corporate Drive
Suite 350
Charlotte, NC 28273
Office: (704) 831-3905 | Mobile: (704) 763-3333

Chris.Fogarty@Versiant.com

-----Original Message-----
From: Quinn Stevenson [mailto:quinn@pronoia-solutions.com] 
Sent: Friday, April 29, 2016 10:45 AM
To: users@activemq.apache.org
Subject: Re: Testing Master Slave on Shared File System

I've done quite a bit of master/slave setups - they've never been an issue as long as I had
a filesystem that supported locking.  I've used NFSv4 and GFSv2 for these setups.

If I'm reading this correctly, you've setup a volume on a SAN and mounted it on both systems
as an ext4 filesystem.  If that's the case, I think that is your issue - ext4 is not a shared
filesystem and it isn't cluster aware.  

Can you try using NFSv4?

> On Apr 29, 2016, at 8:37 AM, Matt Pavlovich <mattrpav@gmail.com> wrote:
> 
> Are both volumes mounted with ext4? Does EXT4 have support for distributed lock sharing?
 Sounds like one server would mount rw and the other would be mounted ro and there aren't
any shared locks.
> 
> I'm not as current on the latest EXT4 features, but do know a cluster-aware filesystem
such as GFSv2 is designed for this type of setup.
> 
> -Matt
> 
> On 4/28/16 3:14 PM, Christopher Fogarty wrote:
>> I have the disk a part of its on vggroup and an lv carved out of that with ext 4
file system on it. This is mounted on both systems and I am able to start active mq fine.
But would feel a lot better validating that only one of the two nodes actually has a lock.
I would love even more to verify that both nodes when started are doing what they should,
which is one has a locked access and the other is in a sort of stand by until the lock is
released.
>> 
>> Hope this makes sense.
>> 
>> Chris Fogarty
>> 
>> 
>> 
>> 
>> 
>> On Thu, Apr 28, 2016 at 12:37 PM -0700, "Matt Pavlovich" <mattrpav@gmail.com<mailto:mattrpav@gmail.com>>
wrote:
>> 
>> Chris-
>> 
>> What file system are you using to share the mount?  The filesystem 
>> would need to support distributed locking (many "shareable 
>> filesystems" don't do this properly.
>> 
>> The other approach is to use the shared filesystem for KahaDB and a 
>> database lease-locker to work around the 
>> most-shared-filesystems-don't-do-locking-properly problem.
>> 
>> -Matt
>> 
>> On 4/28/16 12:34 PM, Christopher Fogarty wrote:
>>> I have set up two servers:
>>> 
>>> Both CENTOS with a shared SAN disk mounted and active on both nodes.
>>> 
>>> I have set up ActiveMQ 5.6
>>> 
>>> I am able to start each with the following configuration
>>> 
>>> <persistenceAdapter>
>>>    <kahaDB directory="/sharedFileSystem/sharedBrokerData"/>
>>> </persistenceAdapter>
>>> 
>>> Each node can and does start, but how can I test, or what do I look for to make
sure that file locking is actually working as described in the http://activemq.apache.org/shared-file-system-master-slave.html
document.  Before putting this into production, I would feel a lot better knowing that only
one of the two nodes is capable of accessing the kahadb.
>>> 
>>> 
>>> Chris Fogarty
>>> 
>>> VP, System Engineering
>>> Versiant Corporation
>>> 3700 Arco Corporate Drive
>>> Suite 350
>>> Charlotte, NC 28273
>>> Office: (704) 831-3905 | Mobile: (704) 763-3333
>>> 
>>> Chris.Fogarty@Versiant.com
>>> 
>>> -----Original Message-----
>>> From: Christopher Fogarty
>>> Sent: Wednesday, April 27, 2016 9:02 AM
>>> To: 'users@activemq.apache.org' <users@activemq.apache.org>
>>> Subject: RE: Running ActiveMQ Broker as different username unable to 
>>> connect via web admin console
>>> 
>>> What Platform? Do you have a firewall running
>>> 
>>> Chris Fogarty
>>> 
>>> VP, System Engineering
>>> Versiant Corporation
>>> 3700 Arco Corporate Drive
>>> Suite 350
>>> Charlotte, NC 28273
>>> Office: (704) 831-3905 | Mobile: (704) 763-3333
>>> 
>>> Chris.Fogarty@Versiant.com
>>> 
>>> -----Original Message-----
>>> From: jboss [mailto:jboss@bcidaho.com]
>>> Sent: Wednesday, April 27, 2016 8:16 AM
>>> To: users@activemq.apache.org
>>> Subject: Re: Running ActiveMQ Broker as different username unable to 
>>> connect via web admin console
>>> 
>>> The web console does not come up at all.   The error that the Chrome gives is
>>> "Connection Refused".  Does not even get to the point of asking for username/password.
>>> 
>>> Thanks,
>>> 
>>> 
>>> 
>>> 
>>> --
>>> View this message in context: 
>>> http://activemq.2283324.n4.nabble.com/Running-ActiveMQ-Broker-as-dif
>>> ferent-username-unable-to-connect-via-web-admin-console-tp4711175p47
>>> 11280.html Sent from the ActiveMQ - User mailing list archive at 
>>> Nabble.com.
>> 
> 


Mime
View raw message