tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ja kub <jjaku...@gmail.com>
Subject Re: backupManager - backup on all nodes and slowly
Date Fri, 07 Feb 2014 21:51:16 GMT
Yes no session is lost, and memory isn't wasted as with delta manager.

With 4 nodes,
when I ask 18080 or 28080 or 38080
node 48080 takes 2 more times than other two backup nodes,
now I stop all

now I tried starting nodes in order 2(8080),3,4,1 and started asking from 2
and not 4 was taking greater load but last started 1 (stil 2 times more
than others)
it looks like the last started tomcat takes greater load
now I stop all

now started in order 3,4,2, 1 was not started, asked 4, node 2 has two
times more sessions than node 3.
now I start node 1, ask 4, and now node 1 takes two time more than other
backup nodes.

now I added another fifth node 58080, started it, and now it takes 40% of
backup sessions, and each of other 3 backup nodes takes 20% of backup
session.

Configuration is the same for all tomcats except from catalina.properties
with port numbers.

/node1$ ls
conf  logs  run.sh  temp  work

node 2, 3, and 4 are copy of node 1,

nodes are started with ./run.sh with the same JVM parameters:

node1$ cat run.sh
export CATALINA_HOME=/szkolenie/tomcat
export CATALINA_BASE=$(pwd)

/szkolenie/tomcat/bin/catalina.sh run


in server.xml:

<Cluster className="org.apache.catalina.ha.tcp.SimpleTcpCluster"
channelSendOptions="6">

                  <Manager
className="org.apache.catalina.ha.session.BackupManager"
                   expireSessionsOnShutdown="false"
                   notifyListenersOnReplication="true"
                   mapSendOptions="6"/>

</Cluster>


BR,
Jakub


On Tue, Feb 4, 2014 at 12:08 PM, Neven Cvetkovic
<neven.cvetkovic@gmail.com>wrote:

> On Tue, Feb 4, 2014 at 4:08 AM, Ja kub <jjakub83@gmail.com> wrote:
>
> > in Manager app I see:
> > after
> > ab  -k -n 100 -p post.txt http://localhost:18080/petclinic/session/fill
> >
> > on 18080 in petclinic 100 sessions
> > on 28080 in petclinic 25 sessions
> > on 38080 in petclinic 25 sessions
> > on 48080 in petclinic 50 sessions     !!!! yes two times more
> >
> > after adding
> > ab  -k -n 900 -p post.txt http://localhost:18080/petclinic/session/fill
> >
> > on 18080 in petclinic 1000 sessions
> > on 28080 in petclinic 250 sessions
> > on 38080 in petclinic 250 sessions
> > on 48080 in petclinic 500 sessions   !!!! yes still two times more
> >
> > Is it not strange ?
> > 18080, 28080, 38080, 48080 have the same configuration.
> >
> > What can be the cause ?
> >
> >
> >
> Jakub,
>
> Well, good thing is that all 100 sessions (+900 later) were
> replicated/backed up (by the BackupManager) to other cluster members that
> host the same application --> 25+25+50 = 100 and 250+250+500 = 1000.
>
> What happens if you repeat the experiment, but without one of the servers,
> e.g. 28080?
>
> Can you send your full cluster configuration? How are you starting you
> servers? Do they start with the same JVM parameters (-Xmx and -Xms)?
>
> Cheers!
> n.
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message