[JBoss JIRA] (GTNPORTAL-2674) Clustering isolation is broken on AS7 - usage of JBoss Cache implementation for JCR/kernel
11:54 a.m.
Michal Vanco created GTNPORTAL-2674:
---------------------------------------
Summary: Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public (Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Fix For: 3.5.0.Beta01
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new cluster
view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new cluster
view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new cluster
view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
11:54 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Michal Vanco updated GTNPORTAL-2674:
------------------------------------
Attachment: probe1.txt
probe2.txt
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta01
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
11:54 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Michal Vanco updated GTNPORTAL-2674:
------------------------------------
Labels: EPP6.0-Test-Dev (was: )
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta01
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
12:39 p.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Marek Posolda commented on GTNPORTAL-2674:
------------------------------------------
Whole this cluster setup is quite tricky because:
- AS7 is using infinispan and jgroups3 for it's own services like session replication
(infinispan subsystem configured in standalone-ha.xml)
- Picketlink IDM is also using infinispan+jgrousp3
- JCR and kernel are still using JBC+jgroups2
And right now, we have separate configuration files for JGroups, which are independent on
AS7 subsystems:
- For JBC+jgroups2 we have gatein.ear/portal.war/WEB-INF/classes/jgroups/gatein-udp.xml
- For infinispan+jgroups3 (actually used by picketlink IDM) we have
gatein.ear/portal.war/WEB-INF/classes/jgroups/gatein-udp-ispn.xml
Actually there are different properties used in those files, so it seems that GateIn needs
to be executed also with different multicast addresses for params
"gatein.jgroups3.udp.mcast_addr" and
"jboss.jgroups.udp_async.mcast_addr" to have multicast addresses properly
isolated.
So maybe something like this will help:
sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b perf12
-Dgatein.jgroups3.udp.mcast_addr=239.4.5.6 -Djboss.jgroups.udp_async.mcast_addr=239.7.8.9
sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b perf13
-Dgatein.jgroups3.udp.mcast_addr=239.6.5.4 -Djboss.jgroups.udp_async.mcast_addr=239.9.8.7
Would be good to verify if it helps. In this case, we may need to simplify configuration
somehow, so that params "gatein.jgroups3.udp.mcast_addr" and
"jboss.jgroups.udp_async.mcast_addr" won't be needed (either improve jgroups
configuration or automatically attach some good values for those params...)
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta01
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
2:51 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Michal Vanco commented on GTNPORTAL-2674:
-----------------------------------------
Hi Marek, adding these two parameters seems to help - gatein.jgroups3.udp.mcast_addr,
jboss.jgroups.udp_async.mcast_addr. Clusters are now separated, confirmed with probe where
all clusters have 1 node only.
Do you think the configuration can be changed so that these parameters are not needed?
And one more important question - is there an option for JCR & kernel that it could
use AS7 clustering subsystem (inf+jgroups3)?
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta01
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
2:20 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Boleslaw Dawidowicz updated GTNPORTAL-2674:
-------------------------------------------
Fix Version/s: 3.5.0.Beta02
(was: 3.5.0.Beta01)
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta02
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
10:49 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Boleslaw Dawidowicz reassigned GTNPORTAL-2674:
----------------------------------------------
Assignee: Marek Posolda
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Assignee: Marek Posolda
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta02
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
10:49 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Thomas Heute commented on GTNPORTAL-2674:
-----------------------------------------
JCR is using JBC+jgroup2.
AFAIK we should document that.
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta02
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
10:07 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Marek Posolda updated GTNPORTAL-2674:
-------------------------------------
Status: Pull Request Sent (was: Open)
Git Pull Request: https://github.com/gatein/gatein-portal/pull/255/
Fixed clustering isolation to work properly without need to add additional options. Now
it's sufficient to use:
./standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
./standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
As multicast address is same for all clusters (infinispan, JBC) and only multicast
differs.
And when same value for parameter "-u" is used, cluster will be fully
established.
I've added also support for TCP cluster (didn't work for AS7)
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Assignee: Marek Posolda
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta02
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
4:53 a.m.
[
https://issues.jboss.org/browse/GTNPORTAL-2674?page=com.atlassian.jira.pl...
]
Boleslaw Dawidowicz updated GTNPORTAL-2674:
-------------------------------------------
Status: Resolved (was: Pull Request Sent)
Resolution: Done
Clustering isolation is broken on AS7 - usage of JBoss Cache
implementation for JCR/kernel
------------------------------------------------------------------------------------------
Key: GTNPORTAL-2674
URL: https://issues.jboss.org/browse/GTNPORTAL-2674
Project: GateIn Portal
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Performance, Security
Affects Versions: 3.4.0.Final
Environment: GateIn master (3.5.0-SNAPSHOT) from 15 Oct
Reporter: Michal Vanco
Assignee: Marek Posolda
Labels: EPP6.0-Test-Dev
Fix For: 3.5.0.Beta02
Attachments: probe1.txt, probe2.txt
I'm hitting an issue with clustering isolation.
To reproduce:
- use two machines
- start first node with: sh standalone.sh -u 239.1.2.3 -c standalone-ha.xml -b ...
- start second node with: sh standalone.sh -u 239.3.2.1 -c standalone-ha.xml -b ...
(keep in mind that these nodes/ do not share gatein data dir, neither datasources - I
don't want to have them connected, I just want to have them started in cluster
profile)
You will see in log:
12:14:56,760 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-5427|1] [perf12-5427, perf13-30570]
12:14:57,171 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-2,null) Received new
cluster view: [perf12-65192|1] [perf12-65192, perf13-9838]
12:14:59,996 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf12-5984|1] [perf12-5984,
perf13-24868]
12:15:00,614 INFO [org.jboss.cache.RPCManagerImpl] (Incoming-1,null) Received new
cluster view: [perf12-53815|1] [perf12-53815, perf13-10269]
If you really want to join the cluster, you should see something like:
12:26:03,894 INFO [org.jboss.as.clustering] (Incoming-1,null) JBAS010225: New cluster
view for partition web (id: 1, delta: 1, merge: false) : [perf13/web, perf12/web]
12:26:03,894 INFO [org.infinispan.remoting.transport.jgroups.JGroupsTransport]
(Incoming-1,null) ISPN000094: Received new cluster view: [perf13/web|1] [perf13/web,
perf12/web]
-u parameter (jboss.default.multicast.address) should take care of clusters isolation
(confirmed by EAP QE).
I've also tried to use probe.sh (from epp 5.2) and I'm attaching output
- probe1.txt is after start of first node (on perf12)
- probe2.txt is after start of second node (on perf13)
Clusters are joined during IDM & JCR initialization.
There could be some out-of-date implementation of JBoss Cache for cluster implementation
of JCR and kernel services :/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
4443
days inactive
4465
days old
9 comments
4 participants
participants (4)
-
Boleslaw Dawidowicz (JIRA) -
Marek Posolda (JIRA)
-
Michal Vanco (JIRA)
-
Thomas Heute (JIRA)