activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gary Tully <gary.tu...@gmail.com>
Subject Re: Failover in 5.2.0
Date Fri, 13 Mar 2009 22:33:07 GMT
The Failover transport layers reconnect logic on top of any of the other
transports.
On NIO, that is typically a broker side option, so the broker will use tcp
using NIO on its transportConnector or listener. A client will still use the
the regular tcp stack.

2009/3/13 Ryan Moquin <fragility2.0@gmail.com>

> That would be fine.. I just want to reestablish the connection if it gets
> dropped (it seems like there are constant network hiccups that cause my
> connection to get lost).  I don't want to lose the connection from
> innactivity either though.  Will it still block if I use the NIO transport?
> Or can I not wrap the NIO transport with the failover transport?
>
> Thanks for your help.
>
> Ryan
>
> On Fri, Mar 13, 2009 at 1:07 PM, Gary Tully <gary.tully@gmail.com> wrote:
>
> > If you are in the same VM, then use the
> > vm:<http://activemq.apache.org/how-should-i-use-the-vm-transport.html
> > >scheme
> > as this will be a lot faster as there is no marshaling overhead.
> >
> > I guess it is worth investigating what is causing your connections to get
> > dropped. Is it inactivity timeout or network issues?
> >
> > With failover, network or connection drop issues will be hidden from your
> > application as the transport will block while it reconnects.
> > There may be situations where you want such issues to filter up to your
> > application to deal with them there, in which case failover: should be
> > avoided or configured to fail after a low number of retries.
> >
> > but it is no harm to use it with a single broker, just be aware that by
> > default it will block with some incremental backoff. check out the
> > reference
> > page for details.
> >
> > 2009/3/13 Ryan Moquin <fragility2.0@gmail.com>
> >
> > > Ahhhhhh ok.. that explains a lot... Too bad I didn't ask this question
> a
> > > lot
> > > sooner since I've been very confused over the last week on why updates
> > > seemed to stop unexpectedly when they never used to.  On the same note,
> > > should I use failover even for interVM connections?  Such as my
> > servicemix
> > > services?  Are there any times it's safe not to use the failover
> > transport?
> > > Or should it always be used unless you have a specific reason you
> > wouldn't
> > > want it to connect?
> > >
> > > Thanks!
> > >
> > > On Fri, Mar 13, 2009 at 12:16 PM, Gary Tully <gary.tully@gmail.com>
> > wrote:
> > >
> > > > you still need to use the failover: scheme in 5.2 when you provide a
> > url
> > > > and
> > > > want failover.
> > > > The change only effects users who do not provide a brokerUrl to the
> > > > ActiveMQConnectionFactory.
> > > > That is the only case where the defaultUrl is used.
> > > >
> > > > 2009/3/13 Ryan Moquin <fragility2.0@gmail.com>
> > > >
> > > > > I've been using ActivbeMQ 5.2 for a while now and am having a
> problem
> > > > where
> > > > > my topic subscribers seem to lose their connections to the ActiveMQ
> > > > broker
> > > > > after around 1,000 messages (it seems like they consistently lose
> > their
> > > > > connection at just over 1,000 messages).  My publishers (SOA
> services
> > > in
> > > > > Servicemix) and my consumers (in a separate VM) are on the same
> > > machine.
> > > >  I
> > > > > can tell my consumers are losing their connections by looking at
> the
> > > > > activemq web console, I can see the connections will just go away
> > while
> > > > my
> > > > > publishers keep sending.  I don't see any error anywhere on my
> > consumer
> > > > > side
> > > > > indicating why the connections were dropped.  Now, assuming some
> > weird
> > > > > network issue (even though they are on the same machine, it seems
> > that
> > > > > network issues will affect local connections) is causing my client
> to
> > > > drop,
> > > > > I thought that ActiveMQ 5.2.0 used failover by default when you use
> a
> > > URL
> > > > > such as:
> > > > >
> > > > > tcp://localhost:61616
> > > > >
> > > > > Since according to the release page it says:
> > > > >
> > > > > "The default ActiveMQConnectionFactory<
> > > > >
> > > >
> > >
> >
> http://activemq.apache.org/maven/activemq-core/apidocs/org/apache/activemq/ActiveMQConnectionFactory.html
> > > > > >brokerUrl
> > > > > now uses the failover
> > > > > transport <
> > > http://activemq.apache.org/failover-transport-reference.html
> > > > >."
> > > > >
> > > > > Am I mistaken on that?  Or do I still need to use the failover URL?
> >  Is
> > > > it
> > > > > common for a consumer to lose a connection like this and using a
> > > failover
> > > > > URL is all I need to put back in?  I never saw this problem when
> > using
> > > > > earlier versions of ActiveMQ with failover.  Since I moved to 5.2,
> I
> > > > > removed
> > > > > the failover part of the URL because I thought 5.2 used it by
> > default.
> > > >  Any
> > > > > clarification on this would be helpful, or even a suggestion on how
> > to
> > > > > troubleshoot the issue.  It seems like the only way I might be able
> > to
> > > > get
> > > > > some log output to help with the problem is by turning on debug
> mode?
> > > > >
> > > > > Thanks for any help!
> > > > >
> > > > > Ryan
> > > > >
> > > >
> > > >
> > > >
> > > > --
> > > > http://blog.garytully.com
> > > >
> > > > Open Source SOA
> > > > http://FUSESource.com
> > > >
> > >
> >
> >
> >
> > --
> > http://blog.garytully.com
> >
> > Open Source SOA
> > http://FUSESource.com
> >
>



-- 
http://blog.garytully.com

Open Source SOA
http://FUSESource.com

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message