tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Yogesh Prajapati <ykp...@gmail.com>
Subject Re: Tomcat 5.5.12 clustering - messages lost under high load
Date Tue, 20 Dec 2005 21:11:51 GMT
On 12/19/05, Peter Rossbach <pr@objektpark.de> wrote:
>
> Hey,
> Am 20.12.2005 um 01:09 schrieb Yogesh Prajapati:
>
> > On 12/18/05, Peter Rossbach <pr@objektpark.de > wrote:
> >>
> >> Hey,
> >>
> >> a)   Servlet Spec say: You must have sticky session when you use
> >> distributable web apps. Session Replication is only used when
> >> primary node
> >> crashed!!
> >
> >
> > I looked into the servlet spec (V 2.4). I did not find anything
> > like "sticky
> > session" or "session replication". I do see the sentence in the spec:
> > "Within an application marked as distributable, all requests that
> > are part
> > of a session must be handled by one Java Virtual Machine1 ("JVM") at a
> > time.", is this what you meant by using "Sticky Session"...if yes,
> > it make
> > sense to me.
> >
> YES!
> >


Great relief to say good bye to such a big problem! many thanks to you! BTW,
meantime I noticed running couple load tests that the problem of "Broken
pipe" exception is gone away (it is fixed).

> b)   When you app don't send a new request before the first is
> > complete:
> >> use pooled mode with waitForAck=true!
> >>        It can work, but....
> >
> > can you please elaborate more on this ...I did not follow what you
> > tried to
> > communicate?
> >
> Pooled mode is a synchronous mode. With waitForAck the response
> thread wait that all backups are received the replication message.


It is interesting to try this option. As you said "It can work, but...." I
understand that by setting "waitForAck=true" will greatly reduce the
throughput therefore not a recommended clustering approach......am I right?

> c)   Rhe reported exception has nothing do with clustering, Seems that
> >> your app send response, before open session. Violate Spec!
> >
> >
> > I think you are right, it is not a cluster exception But I don't
> > see the
> > exception for every request, it appears randomly not sure what is
> > causing
> > it, also I dont know how my JSP based app is violating the spec.
> > All my JSPs
> > are doing is to use "<jsp:include..../>" with two level nesting. E.g.
> >
>
> Why you set JspWriter out=null?? Thats wrong!
> Implement the response header setting inside a servlet filter is better!


I kind of debated myself about this sometime ago, but since JSP pre-compiler
never gave any warning/error (or ignored it) it never caught my attention. I
am crazy at the moment after having gone through it so much that I am unable
to control every single little thing. Thanks very much for pointing that
out.

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