plugin - marking clone as down - Websphere
This is a discussion on plugin - marking clone as down - Websphere ; Hi all,
I have a problem with the plugin-cfg.xml
We have 2 physical servers was 5.1.x these servers are clustered and on each
are 4 clones running.
if a clone goes down I can see this in the plugin log ...
-
plugin - marking clone as down
Hi all,
I have a problem with the plugin-cfg.xml
We have 2 physical servers was 5.1.x these servers are clustered and on each
are 4 clones running.
if a clone goes down I can see this in the plugin log at the webserver
machine, a message like this is logged
[Tue May 22 10:09:57 2007] 00007a65 b7bca6b0 - ERROR: ws_common:
websphereExecute: Failed to read from a new stream; App Server may have
gone down during read
[Tue May 22 10:09:57 2007] 00007a65 b7bca6b0 - ERROR: ws_server:
serverSetFailoverStatus: Marking ws12_WebSphere_Portal-ws12-4 down
[Tue May 22 10:09:57 2007] 00007a65 b7bca6b0 - ERROR: ws_common:
websphereHandleRequest: Failed to execute the transaction
to 'ws12_WebSphere_Portal-ws12-4'on host 'ws12.vitanet.de'; will try
another one
thats ok so far, but after a few seconds a new request is routed to this
clone.
As far as I understand, the plugin should wait at least the time specified
in RetryInterval="300" thats in my case 5 minutes.
Does anybody know or has an idea why the webserver is trying to route a
request to the clone?
TIA
Christian
-
Re: plugin - marking clone as down
christian zimmermann wrote:
> Hi all,
> Does anybody know or has an idea why the webserver is trying to route a
> request to the clone?
What webserver are you using? if it's an Apache derivative that us
running multiple processes, each webserver process needs to discover
that the server is down.
-
Re: plugin - marking clone as down
Paul Ilechko wrote:
> What webserver are you using? if it's an Apache derivative that us
> running multiple processes, each webserver process needs to discover
> that the server is down.
We are using the original apache from apache.org version 2.0.58
with prefork mpm.
Should the worker mpm solve the problem ?
TIA
christian
-
Re: plugin - marking clone as down
christian zimmermann wrote:
> Paul Ilechko wrote:
>
>
>> What webserver are you using? if it's an Apache derivative that us
>> running multiple processes, each webserver process needs to discover
>> that the server is down.
>
> We are using the original apache from apache.org version 2.0.58
> with prefork mpm.
>
> Should the worker mpm solve the problem ?
It's not really a problem, it's just the nature of the beast. You need
to use a fully threaded model if you want a single discovery to be
sufficient.
-
Re: plugin - marking clone as down
"Paul Ilechko" schrieb im Newsbeitrag
news:f31j95$3bc76$1@news.boulder.ibm.com...
> christian zimmermann wrote:
>> Paul Ilechko wrote:
>>
>>
>>> What webserver are you using? if it's an Apache derivative that us
>>> running multiple processes, each webserver process needs to discover
>>> that the server is down.
>>
>> We are using the original apache from apache.org version 2.0.58
>> with prefork mpm.
>>
>> Should the worker mpm solve the problem ?
>
> It's not really a problem, it's just the nature of the beast. You need to
> use a fully threaded model if you want a single discovery to be
> sufficient.
Its that the way it should work ? there are a lot of requests that not get
served.
I thought the goal of the plugin is to provide a failover if a clone is not
available and blocks further requests to this clone,
for a specific time.
Is there another way to achive this goal?
Is the IBM HttpServer a better choice for this job ?
Thanks for your time
Christian
-
Re: plugin - marking clone as down
christian zimmermann wrote:
> "Paul Ilechko" schrieb im Newsbeitrag
> news:f31j95$3bc76$1@news.boulder.ibm.com...
>> christian zimmermann wrote:
>>> Paul Ilechko wrote:
>>>
>>>
>>>> What webserver are you using? if it's an Apache derivative that us
>>>> running multiple processes, each webserver process needs to discover
>>>> that the server is down.
>>> We are using the original apache from apache.org version 2.0.58
>>> with prefork mpm.
>>>
>>> Should the worker mpm solve the problem ?
>> It's not really a problem, it's just the nature of the beast. You need to
>> use a fully threaded model if you want a single discovery to be
>> sufficient.
>
> Its that the way it should work ? there are a lot of requests that not get
> served.
> I thought the goal of the plugin is to provide a failover if a clone is not
> available and blocks further requests to this clone,
> for a specific time.
> Is there another way to achive this goal?
> Is the IBM HttpServer a better choice for this job ?
The point is that every HTTP process has to discover that the server is
down. If your webserver is a single process running lots of threads,
this only happens once. If your webserver spawns lots of processes, then
each process will discover it individually. IBM HTTP Server is a variant
of Apache. It's really a matter of how you configure it. You want to use
a threaded model with few processes.
