Backup Server is not starting as live server when artemis is restarted

[Doug Whitfield <dw...@perforce.com.INVALID>]

Hi folks,

Artemis 2.22.0 is in use. It is not clear to me if there is a separate list for Artemis or not.

In this case, slave started running as live server as master server went down. During the troubleshooting , we restarted the artemis on the slave and it started as a backup server instead of live server. As we couldn't recover the active server, MQ services were not running on both master and slave.

Is there any configuration change that needs to be done whenever the master server is not available and backup server should be running as live server even if there is a restart on the backup server.

Is there any additional information I should send?

Best Regards,
Doug Whitfield


This e-mail may contain information that is privileged or confidential. If you are not the intended recipient, please delete the e-mail and any attachments and notify us immediately.

Re: Backup Server is not starting as live server when artemis is restarted

[Justin Bertram <jb...@apache.org>]

If you're using replication then there is some nuance to how the backup
will behave when it is started in the absence of its primary. The first
thing to note is that the broker configured as a backup doesn't know if the
journal data it has is the most up-to-date data. For example, if the backup
was shutdown while the primary was still active and then the primary is
shutdown and then the backup is started. In this case the backup will not
have all the data from when it was down. Activating the backup fully in
this case could result in apparently lost messages and duplicate messages
as well.

Therefore, if you want a replicated backup to activate fully in the absence
of its corresponding primary then you need to change the configuration in
broker.xml so that it's a master.

I'm fairly certain that this behavior is different if you're using the new
pluggable quorum implementation with ZooKeeper as ZooKeeper acts as a third
party which can inform the backup whether or not it should start in this
kind of situation.

Also, it's worth noting that this doesn't happen when using a shared store
either since there's only one instance of the data (i.e. on the shared
store) rather than a copy on each broker.


Justin


On Tue, Mar 28, 2023 at 11:56 AM Doug Whitfield
<dw...@perforce.com.invalid> wrote:

> Hi folks,
>
> Artemis 2.22.0 is in use. It is not clear to me if there is a separate
> list for Artemis or not.
>
> In this case, slave started running as live server as master server went
> down. During the troubleshooting , we restarted the artemis on the slave
> and it started as a backup server instead of live server. As we couldn't
> recover the active server, MQ services were not running on both master and
> slave.
>
> Is there any configuration change that needs to be done whenever the
> master server is not available and backup server should be running as live
> server even if there is a restart on the backup server.
>
> Is there any additional information I should send?
>
> Best Regards,
> Doug Whitfield
>
>
> This e-mail may contain information that is privileged or confidential. If
> you are not the intended recipient, please delete the e-mail and any
> attachments and notify us immediately.
>
>