You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@activemq.apache.org by khandelwalanuj <an...@gmail.com> on 2017/04/26 05:51:43 UTC
ActiveMQ JDBC with journaling enabled
Hi,
I was evaluating ActiveMQ's JDBC based persistent store with journaling
enabled. Have couple of doubts:
1. I was reading
https://access.redhat.com/documentation/en-US/Fuse_Message_Broker/5.4/html/Clustering_Guide/files/Failover-MasterSlave-JDBC.html
which says that the journaled JDBC store is incompatible with the JDBC
master/slave failover pattern. Does it mean client automatic failover won't
happen here ? What does it exactly mean ?
2. If master slave is properly supported, is persistent message loss is
possible in journaled JDBC. Below is one scenario:
" If both the brokers are running on two different hosts and each one
has it's own journal store(activemq-data) and both shares the same backend
DB (let's Say postgres). In journaled jdbc if both producer and consumer are
up, it won't store the message in backend DB which are received between a
single checkpoint. It will keep it in journal store(activemq-data). Now
let's say at this point to time if master goes down, does the messages which
are sent to the master but not yet delivered to the consumer, are lost as
new master will start from it's own journal and same same backend DB, but
since backend DB doesn't have the message. So is it lost ? "
Thanks,
Anuj
--
View this message in context: http://activemq.2283324.n4.nabble.com/ActiveMQ-JDBC-with-journaling-enabled-tp4725234.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
Re: ActiveMQ JDBC with journaling enabled
Posted by Martyn Taylor <mt...@redhat.com>.
My answer was based on ActiveMQ Artemis. Sorry for the confusion.
On Wed, Apr 26, 2017 at 1:56 PM, khandelwalanuj <
anuj.cool.khandelwal@gmail.com> wrote:
> Yes I was talking about ActiveMQ 5.x
>
> Can you also clarify about the first point where it says that journaled
> JDBC
> store is incompatible with the JDBC master/slave. What does it exactly mean
> ? Link which says this:
> https://access.redhat.com/documentation/en-US/Fuse_
> Message_Broker/5.4/html/Clustering_Guide/files/
> Failover-MasterSlave-JDBC.html
>
> Thanks,
> Anuj
>
>
>
>
>
> --
> View this message in context: http://activemq.2283324.n4.
> nabble.com/ActiveMQ-JDBC-with-journaling-enabled-tp4725234p4725239.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>
Re: ActiveMQ JDBC with journaling enabled
Posted by Tim Bain <tb...@alumni.duke.edu>.
I interpret that document to be written in the context of a requirement
that the broker not lose persistent messages in the event of a restart or
failover, since those are the requirements of most ActiveMQ users using a
persistence store. (Typically, either you need 100% persistence and you use
a persistence store or you don't need any and you just use the memory store
because it's faster; few people want some-but-not-necessarily-all behavior,
which is what you'd get in the setup you're describing.) So when they say
"it's necessary" or "you must" (never "it's incompatible" for the record),
I believe the unwritten rest of the statement is "because if you did, you'd
lose messages."
But since that document is written, published, and maintained by the Fuse
developers, all of what I just wrote is just guesswork, and if my guesswork
isn't sufficient and you want definitive answers about what they meant by
what they wrote, you'll need to ask them.
Tim
On Apr 26, 2017 7:08 AM, "khandelwalanuj" <an...@gmail.com>
wrote:
> Yes I was talking about ActiveMQ 5.x
>
> Can you also clarify about the first point where it says that journaled
> JDBC
> store is incompatible with the JDBC master/slave. What does it exactly mean
> ? Link which says this:
> https://access.redhat.com/documentation/en-US/Fuse_
> Message_Broker/5.4/html/Clustering_Guide/files/
> Failover-MasterSlave-JDBC.html
>
> Thanks,
> Anuj
>
>
>
>
>
> --
> View this message in context: http://activemq.2283324.n4.
> nabble.com/ActiveMQ-JDBC-with-journaling-enabled-tp4725234p4725239.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>
Re: ActiveMQ JDBC with journaling enabled
Posted by khandelwalanuj <an...@gmail.com>.
Yes I was talking about ActiveMQ 5.x
Can you also clarify about the first point where it says that journaled JDBC
store is incompatible with the JDBC master/slave. What does it exactly mean
? Link which says this:
https://access.redhat.com/documentation/en-US/Fuse_Message_Broker/5.4/html/Clustering_Guide/files/Failover-MasterSlave-JDBC.html
Thanks,
Anuj
--
View this message in context: http://activemq.2283324.n4.nabble.com/ActiveMQ-JDBC-with-journaling-enabled-tp4725234p4725239.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
Re: ActiveMQ JDBC with journaling enabled
Posted by Tim Bain <tb...@alumni.duke.edu>.
Anuj and Martyn, were your question and your answer about ActiveMQ 5.x or
Artemis? I think the question was about 5.x and Martyn's answer was about
Artemis. If so, see a 5.x response below.
I don't believe that there is any reason other than the potential for
message loss that you described (correctly) in #2 that would make the page
you linked to in #1 say that you need to use non-journaled JDBC. If your
use case can support the loss of messages since the last checkpoint, I
expect that you can make it work. If you choose to pursue it, you may want
to use pluggable storage lockers (
http://activemq.apache.org/pluggable-storage-lockers.html) for master
elections if whatever is built into the journaled JDBC code doesn't work; I
have no idea how locks would interact with journaling.
Tim
On Apr 26, 2017 3:16 AM, "Martyn Taylor" <mt...@redhat.com> wrote:
> Hi
>
> On Wed, Apr 26, 2017 at 6:51 AM, khandelwalanuj <
> anuj.cool.khandelwal@gmail.com> wrote:
>
> > Hi,
> >
> > I was evaluating ActiveMQ's JDBC based persistent store with journaling
> > enabled. Have couple of doubts:
> >
> > 1. I was reading
> > https://access.redhat.com/documentation/en-US/Fuse_
> > Message_Broker/5.4/html/Clustering_Guide/files/
> > Failover-MasterSlave-JDBC.html
> > which says that the journaled JDBC store is incompatible with the JDBC
> > master/slave failover pattern. Does it mean client automatic failover
> won't
> > happen here ? What does it exactly mean ?
> >
> Right now we don't have support for automated HA using the JDBC store. The
> JDBC store is a relatively new feature and the focus thus far has been on
> stability. However, adding HA is something we have planned.
>
> >
> > 2. If master slave is properly supported, is persistent message loss is
> > possible in journaled JDBC. Below is one scenario:
> > " If both the brokers are running on two different hosts and each one
> > has it's own journal store(activemq-data) and both shares the same
> backend
> > DB (let's Say postgres). In journaled jdbc if both producer and consumer
> > are
> > up, it won't store the message in backend DB which are received between a
> > single checkpoint. It will keep it in journal store(activemq-data). Now
> > let's say at this point to time if master goes down, does the messages
> > which
> > are sent to the master but not yet delivered to the consumer, are lost as
> > new master will start from it's own journal and same same backend DB, but
> > since backend DB doesn't have the message. So is it lost ? "
> >
>
> I'm not 100% following the scenario here. But in short, if you are using
> persistent messages (or an appropriate quality of service). You'll never
> get in a situation where you have messages lost. This is one of the
> fundamental requirements of the broker.
>
> >
> > Thanks,
> > Anuj
> >
> >
> >
> > --
> > View this message in context: http://activemq.2283324.n4.
> > nabble.com/ActiveMQ-JDBC-with-journaling-enabled-tp4725234.html
> > Sent from the ActiveMQ - User mailing list archive at Nabble.com.
> >
>
Re: ActiveMQ JDBC with journaling enabled
Posted by Martyn Taylor <mt...@redhat.com>.
Hi
On Wed, Apr 26, 2017 at 6:51 AM, khandelwalanuj <
anuj.cool.khandelwal@gmail.com> wrote:
> Hi,
>
> I was evaluating ActiveMQ's JDBC based persistent store with journaling
> enabled. Have couple of doubts:
>
> 1. I was reading
> https://access.redhat.com/documentation/en-US/Fuse_
> Message_Broker/5.4/html/Clustering_Guide/files/
> Failover-MasterSlave-JDBC.html
> which says that the journaled JDBC store is incompatible with the JDBC
> master/slave failover pattern. Does it mean client automatic failover won't
> happen here ? What does it exactly mean ?
>
Right now we don't have support for automated HA using the JDBC store. The
JDBC store is a relatively new feature and the focus thus far has been on
stability. However, adding HA is something we have planned.
>
> 2. If master slave is properly supported, is persistent message loss is
> possible in journaled JDBC. Below is one scenario:
> " If both the brokers are running on two different hosts and each one
> has it's own journal store(activemq-data) and both shares the same backend
> DB (let's Say postgres). In journaled jdbc if both producer and consumer
> are
> up, it won't store the message in backend DB which are received between a
> single checkpoint. It will keep it in journal store(activemq-data). Now
> let's say at this point to time if master goes down, does the messages
> which
> are sent to the master but not yet delivered to the consumer, are lost as
> new master will start from it's own journal and same same backend DB, but
> since backend DB doesn't have the message. So is it lost ? "
>
I'm not 100% following the scenario here. But in short, if you are using
persistent messages (or an appropriate quality of service). You'll never
get in a situation where you have messages lost. This is one of the
fundamental requirements of the broker.
>
> Thanks,
> Anuj
>
>
>
> --
> View this message in context: http://activemq.2283324.n4.
> nabble.com/ActiveMQ-JDBC-with-journaling-enabled-tp4725234.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>