[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17033869#comment-17033869
]
Justin Bertram commented on ARTEMIS-2165:
-
[~ryeats], I think that's a good idea. I've kicked off a "DISCUSS" thread on
the ActiveMQ developers list related to this.
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17029151#comment-17029151
]
Ryan Yeats commented on ARTEMIS-2165:
-
[~jbertram] Would it be better to remove colocation as an option for
replication in the mean time until this feature gets more love since we see a
lot of emails in the user email list over the years of people wasting time
trying to use it and there is not an actual working example in which it can
failover successfully.
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16968429#comment-16968429
]
Bob Mitchell commented on ARTEMIS-2165:
---
We have also run into this and find it to be a significant problem.
My research has shown that the major problem is the following sequence:
# Server comes up and requests a backup server
# It finds a server and attaches to it.
# The backup server goes down.
At this point, the original server will never request to be backed up again.
What is worse, is while you can get it to start backing up again by stopping
and starting it, if that server is backing up another server you just pushed
the problem to the other server. In some cases, this may mean you cannot get
all of the servers being backed up without bringing down all of the servers and
then restarting them, which is not acceptable in our use case.
On a secondary note, we are finding it necessary to set max-backups > 1 if
there are an odd number of servers in the cluster. The problem is that the
first two servers to start up grab each other as backups and the third server
is left without a server offering a backup unless at least one of them allows
multiple backups. The only way around this would be to use group-name settings
to designate which slave backs up each master (i.e. use one group name for each
master and assign each slave one of the other server's master group name).
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16867097#comment-16867097
]
Dan Langford commented on ARTEMIS-2165:
---
FWIW we are experiencing similar issues. A test case was written for our
problem. It didnt get much traction over in the mailing list but maybe it will
prove useful to somebody over here. The test case is here:
[https://github.com/SethPyle376/colocated-scaledown-problem/blob/master/src/main/java/org/apache/activemq/artemis/jms/example/ColocatedFailoverScaledownFailbackTest.java]
it shows that when a server in a colocated ha-policy is shut down and then
restarted that server fails to (re-)obtain a backup on another server. it
demonstrates that by producing a message, crashing the server, and showing that
the messages didnt get backed up and scaled down... which is an opportunity for
message loss.
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16701012#comment-16701012
]
Justin Bertram commented on ARTEMIS-2165:
-
How are you identifying the live servers and the backup that they're paired
with? Could you work up a simple test-case (e.g. by modifying one of the
examples we ship with the broker) to demonstrate?
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16686610#comment-16686610
]
Dmitry Smeliansky commented on ARTEMIS-2165:
Is it possible to get some update for this one?
As I mentioned in the ticket description, this exactly the same question was
asked multiple times in stackoverflow but no clear answer was provided.
This is one of our decision makers about using Artemis as a queue
infrastructure of our product.
Thanks
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16682759#comment-16682759
]
Dmitry Smeliansky commented on ARTEMIS-2165:
Thanks for the answer.
We did try the configuration you've suggested (before opening the ticket), but
the result was the same. We also tried to define cluster and group names for
the master/slave.
The latest HA-policy configuration we used are the following:
For BROKER1:
{code:java}
100
BROKER2
true
1
-1
5000
artemis-cluster
BROKER1
true
artemis-cluster
BROKER2
true
true
-1
{code}
And for BROKER2:
{code:java}
100
BROKER1
true
1
-1
5000
artemis-cluster
BROKER2
true
artemis-cluster
BROKER1
true
true
-1
{code}
The result is always the same: before restarting one of the nodes we have 4
brokers - 2 live and 2 backup, after one node is restarted we have 3 lives and
no backups
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16681825#comment-16681825
]
Justin Bertram commented on ARTEMIS-2165:
-
I think you can the behavior you're looking for by using this:
{code:xml}
true
true
{code}
Also, I would recommend against configuring {{max-backups}} > 1 for a couple of
reasons:
* Multiple backups preclude the expected fail-back behavior.
* A live only replicates to a single back-up at a time.
* Multiple failures results in an increasing number of brokers running on a
single machine which is likely to overwhelm the hardware (assuming there is
load sufficient to warrant a cluster that size).
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> We use colocated HA-policy and trying to achieve the following:
> # We have 2 servers and want each one of them will hold a) 1 master b) 1
> slave for the master of each other
> # When we start the servers we have the above as we wanted
> # When one of the server is going down, the slave on the second becomes a
> live master - till now everything is ok
> # When the second server (which was down in step 3) is up, the slave from
> server 1 remains a live master and a slave for master 1 is not created in
> server 2.
> All we want is : how the system should be configured, so after the step 4 we
> will have the exact topology as after the step 1.
> # Start server 1 and server 2: server 1 has 1 master and 1 slave, server 2
> has one master and one slave
> # Restart server 2: server 1 has one master and one slave, server 2 has one
> master and one slave
>
>
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (ARTEMIS-2165) Not having backup available after restart in colocated configuration
[
https://issues.apache.org/jira/browse/ARTEMIS-2165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16678231#comment-16678231
]
Justin Bertram commented on ARTEMIS-2165:
-
[~dmitrysm2000], instead of linking to Stack Overflow please put a full
description of the issue in the issue {{Description}} including steps to
reproduce.
> Not having backup available after restart in colocated configuration
>
>
> Key: ARTEMIS-2165
> URL: https://issues.apache.org/jira/browse/ARTEMIS-2165
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: Broker
>Affects Versions: 2.6.0
>Reporter: Dmitry Smeliansky
>Priority: Major
>
> This issue was asked multiple times in stackoverflow, but was never answered.
> When using a co-located HA policy, the backups, maintained by some node are
> not restored after this node is restarted (not on this node and not on any
> other).
> [https://stackoverflow.com/questions/52839706/artemis-2-6-0-colocated-ha-issues]
> [https://stackoverflow.com/questions/52000662/artemis-colocated-ha-policy]
> Our configuration was as the following:
> {code:java}
>
>
>
> 100
> true
> 2
> -1
>
> 5000
>
>
>
>
> {code}
> Thanks
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
