2:06 p.m.
*Which properties do I have to set (I guess in mod-cluster-jboss.beans
in HAModClusterConfig) to remove crashed workers from http's tables ?*
I want httpd (mod-cluster) to periodically try to connect to a worker
and remove the worker is the connection is dead.
In mod_jk, I guess (according to [1]) I'd have to set ping_mode=I, with
ping_timeout being the interval and connection_ping_interval to a value
in secs after which a conn will be checked if idle.
I don't know which props I have to use with mod_cluster: nodeTimeout ?
[2] I guess that's only for open sockets though...
[1] http://tomcat.apache.org/connectors-doc/reference/workers.html
[2] http://jboss.org/mod_cluster/native/mod_jk.html
--
Bela Ban
Lead JGroups / Clustering Team
JBoss
2:15 p.m.
In HAModClusterConfig, it is sufficient to set
workerTimeout=5000 (was -1)
nodeTimeout=5000 (didn't exist)
?
When using multiple domains should I also set masterPerDomain=true ?
What's the advantage of this setting ? Isn't this necessary anyway,
because if clusters don't span domains, masters can't span domains either ?
Bela Ban wrote:
*Which properties do I have to set (I guess in
mod-cluster-jboss.beans
in HAModClusterConfig) to remove crashed workers from http's tables ?*
I want httpd (mod-cluster) to periodically try to connect to a worker
and remove the worker is the connection is dead.
In mod_jk, I guess (according to [1]) I'd have to set ping_mode=I, with
ping_timeout being the interval and connection_ping_interval to a value
in secs after which a conn will be checked if idle.
I don't know which props I have to use with mod_cluster: nodeTimeout ?
[2] I guess that's only for open sockets though...
[1] http://tomcat.apache.org/connectors-doc/reference/workers.html
[2] http://jboss.org/mod_cluster/native/mod_jk.html
--
Bela Ban
Lead JGroups / Clustering Team
JBoss
4:47 p.m.
On 08/13/2009 02:06 PM, Bela Ban wrote:
*Which properties do I have to set (I guess in
mod-cluster-jboss.beans
in HAModClusterConfig) to remove crashed workers from http's tables ?*
You can't... The idea is that the cluster logic will do it by removing
the node.
Cheers
Jean-Frederic
4:56 p.m.
But this works in mod_jk doesn't it ?
Besides, when we have *non-clustered* workers, there is no HA singleton
telling httpd to remove the crashed worker, so httpd has to do it itself.
I mean, why can't httpd simply remove a worker W is the socket
connection to W is closed (by W crashing) ?
jean-frederic clere wrote:
On 08/13/2009 02:06 PM, Bela Ban wrote:
> *Which properties do I have to set (I guess in mod-cluster-jboss.beans
> in HAModClusterConfig) to remove crashed workers from http's tables ?*
You can't... The idea is that the cluster logic will do it by removing
the node.
Cheers
Jean-Frederic
--
Bela Ban
Lead JGroups / Clustering Team
JBoss
5:13 p.m.
On 08/13/2009 04:56 PM, Bela Ban wrote:
But this works in mod_jk doesn't it ?
mod_jk has a static configuration it is not going to remove something
from the configuration.
Besides, when we have *non-clustered* workers, there is no HA singleton
telling httpd to remove the crashed worker, so httpd has to do it itself.
I mean, why can't httpd simply remove a worker W is the socket
connection to W is closed (by W crashing) ?
First an entry corresponding shouldn't disturb mod_cluster, see
MODCLUSTER-92 about telling it is broken. Of course it would possible to
remove broken workers after a while, but I think it must be a
switch-able option and the while should be a parameter... Please create
a JIRA.
Cheers
Jean-Frederic
jean-frederic clere wrote:
> On 08/13/2009 02:06 PM, Bela Ban wrote:
>> *Which properties do I have to set (I guess in mod-cluster-jboss.beans
>> in HAModClusterConfig) to remove crashed workers from http's tables ?*
>
> You can't... The idea is that the cluster logic will do it by removing
> the node.
>
> Cheers
>
> Jean-Frederic
6:18 p.m.
jean-frederic clere wrote:
On 08/13/2009 04:56 PM, Bela Ban wrote:
> But this works in mod_jk doesn't it ?
mod_jk has a static configuration it is not going to remove something
from the configuration.
>
> Besides, when we have *non-clustered* workers, there is no HA singleton
> telling httpd to remove the crashed worker, so httpd has to do it
> itself.
>
> I mean, why can't httpd simply remove a worker W is the socket
> connection to W is closed (by W crashing) ?
First an entry corresponding shouldn't disturb mod_cluster, see
MODCLUSTER-92 about telling it is broken. Of course it would possible
to remove broken workers after a while, but I think it must be a
switch-able option and the while should be a parameter... Please
create a JIRA.
I think what should happen is
- httpd periodically tests the connection to worker W
- httpd detects connection loss to W
- httpd removes W from its worker tables, so requests are not dispatched
to W
- httpd also starts a timer, going off in configurable intervals, which
tests W again. If W comes up again, httpd adds it again
Actually, I think the last point is not necessary, because when W is
started again, mod-cluster/jboss will tell httpd ! We should actually
*not* implement the last point because W might never get started again !
WDYT ?
--
Bela Ban
Lead JGroups / Clustering Team
JBoss
6:35 p.m.
On 08/13/2009 06:18 PM, Bela Ban wrote:
jean-frederic clere wrote:
> On 08/13/2009 04:56 PM, Bela Ban wrote:
>> But this works in mod_jk doesn't it ?
>
> mod_jk has a static configuration it is not going to remove something
> from the configuration.
>
>>
>> Besides, when we have *non-clustered* workers, there is no HA singleton
>> telling httpd to remove the crashed worker, so httpd has to do it
>> itself.
>>
>> I mean, why can't httpd simply remove a worker W is the socket
>> connection to W is closed (by W crashing) ?
>
> First an entry corresponding shouldn't disturb mod_cluster, see
> MODCLUSTER-92 about telling it is broken. Of course it would possible
> to remove broken workers after a while, but I think it must be a
> switch-able option and the while should be a parameter... Please
> create a JIRA.
I think what should happen is
- httpd periodically tests the connection to worker W
Via cping/cpong but only if the worker is not active (no requests for a
while).
- httpd detects connection loss to W
- httpd removes W from its worker tables, so requests are not dispatched
to W
As soon as the worker W is marked in error the request are not forwarded
to it. (That is already the since the very first beta of mod_cluster).
So removing would be removing a node, virtualhosts and contexts.
- httpd also starts a timer, going off in configurable intervals,
which
tests W again. If W comes up again, httpd adds it again
NO:
- We can't unremove stuff.
- May be the customer has changed his configuration and doesn't want
that httpd to proxy this node.
- the JAVA part will create it anyway (your comment below).
Actually, I think the last point is not necessary, because when W is
started again, mod-cluster/jboss will tell httpd ! We should actually
*not* implement the last point because W might never get started again !
Yep ;-)
Cheers
Jean-Frederic
WDYT ?
6:50 p.m.
jean-frederic clere wrote:
> I think what should happen is
> - httpd periodically tests the connection to worker W
Via cping/cpong but only if the worker is not active (no requests for
a while).
How do we configure this ? I know how to do it in mod_jk, but haven't
found the equivalent config in mod-cluster !
> - httpd detects connection loss to W
> - httpd removes W from its worker tables, so requests are not dispatched
> to W
As soon as the worker W is marked in error the request are not
forwarded to it. (That is already the since the very first beta of
mod_cluster).
So removing would be removing a node, virtualhosts and contexts.
Yes. But when this happens in mod-cluster, e.g. by killing -9 a node, it
takes almost a minute to fail over. Which config governs this timeout ?
How do I do this ? Help please ? URGENT !!! LOL :-)
--
Bela Ban
Lead JGroups / Clustering Team
JBoss
8:38 a.m.
On 08/13/2009 06:50 PM, Bela Ban wrote:
jean-frederic clere wrote:
>> I think what should happen is
>> - httpd periodically tests the connection to worker W
> Via cping/cpong but only if the worker is not active (no
requests for
> a while).
How do we configure this ? I know how to do it in mod_jk, but haven't
found the equivalent config in mod-cluster !
Right wee don't have that in mod_cluster.
>> - httpd detects connection loss to W
>> - httpd removes W from its worker tables, so requests are not dispatched
>> to W
> As soon as the worker W is marked in error the request are
not
> forwarded to it. (That is already the since the very first beta of
> mod_cluster).
> So removing would be removing a node, virtualhosts and contexts.
Yes. But when this happens in mod-cluster, e.g. by killing -9 a node, it
takes almost a minute to fail over. Which config governs this timeout ?
Must be in httpd and probably with a default value being n time the
cping/cpong timeout.
How do I do this ? Help please ? URGENT !!! LOL :-)
I can't make it for 1.0.2 :-).
All the logic pieces are in the httpd code part, but it always tricky in
httpd to send and receive outside a real response/request. Open a JIRA
for 1.1.x.
Cheers
Jean-Frederic
9:08 a.m.
jean-frederic clere wrote:
I can't make it for 1.0.2 :-).
No problem. I just want to identify all issues we need to fix for 1.1 or
later
All the logic pieces are in the httpd code part, but it always tricky
in httpd to send and receive outside a real response/request. Open a
JIRA for 1.1.x.
https://jira.jboss.org/jira/browse/MODCLUSTER-97
--
Bela Ban
Lead JGroups / Clustering Team
JBoss
10:28 a.m.
Now I'm confused: I ran the following scenario, and it worked !
* Nodes A and B
* Session is on B
* kill -9 of B
* *Access the session, it fails over to A immediately !*
* (mod_cluster_manager still shows both nodes, but this is a known
issue)
So httpd seems to remove a worker W if it cannot establish a connection
to W ? I thought we came to the conclusion that this wasn't the case ?
--
Bela Ban
Lead JGroups / Clustering Team
JBoss
11:31 a.m.
On 08/14/2009 10:28 AM, Bela Ban wrote:
Now I'm confused: I ran the following scenario, and it worked !
* Nodes A and B
* Session is on B
* kill -9 of B
* *Access the session, it fails over to A immediately !*
* (mod_cluster_manager still shows both nodes, but this is a known
issue)
So httpd seems to remove a worker W if it cannot establish a connection
to W ? I thought we came to the conclusion that this wasn't the case ?
Probably because remove for you means STOPPED and NOTOK (Not usable for
any reason, like a result of a kill -9) and remove for me means remove
from the shared tables (from all the tables, node, hosts and contexts).
Cheers
Jean-Frederic
5369
days inactive
5370
days old
mod_cluster-dev@lists.jboss.org
11 comments
2 participants
participants (2)
-
Bela Ban -
jean-frederic clere