[ARTEMIS-2353] STOMP message loss during failover

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Not A Problem
• Affects Version/s: 2.6.4
• Component/s: Broker, STOMP

Environment

REDHAT_BUGZILLA_PRODUCT="Red Hat Enterprise Linux 7"
REDHAT_BUGZILLA_PRODUCT_VERSION=7.5
REDHAT_SUPPORT_PRODUCT="Red Hat Enterprise Linux"
REDHAT_SUPPORT_PRODUCT_VERSION="7.5"

• Assignee: Justin Bertram
• Reporter: Gaurav
• Created: 23/May/19 13:11
• Updated: 26/Jun/19 20:06
• Resolved: 26/Jun/19 20:05

Description

We have single instance of Artemis Live server (2.6.4 version) backed up by Backup server.  Please refer attachments for both Live-Backup broker configuration.

We are using stomp protocol for messaging (port 61613).

During failover of Live server, backup server is taking control of processing all incoming requests so switching from Live->Backup server is smooth.

But we are seeing message loss (around 2 to 5 messages) exactly during fail over time. These messages were successfully published from client side and no exception received from broker by this time. After publishing of these messages, connectionLost exception was thrown during fail over which our application is correctly handling to shift incoming messages load to other node.

It seems Live instance could not persist few messages on Journal file system as it got wiped out from in memory session. Major issue is that Client is not getting notified through any exception that message could not persist on broker journal.

 We are looking for broker configuration through which client call can be blocked until published message is not persisted on Journal.

Any pointer on this issue is appreciated!