You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tomcat.apache.org by Remy Maucherat <re...@apache.org> on 2002/06/26 13:14:13 UTC
[4.1.6] New milestone release soon
There are only a few issues remaining:
- Updating JNDI resources with the admin webapp is not dynamic (for
reasons currently beyond my understanding). Doing a stop/start on the
context allows to pick up the changes, so the bug is only minor.
- Nacho's IIS issues with JK 2.
- Costin's bug with Jasper 2.
None of these are showstoppers IMO, but it would be best to have them
fixed before the release is tagged (the objective being to get back to
beta status). By friday maybe ?
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
> Here is a bit more information on why I haven't ruled out JSTL or 1.2
> JSP tags as a possible cause.
>
> 1. the standard examples all run fine and do not exhibit a memory leak,
> including the simpletag example.
> 2. the other person experiencing memory leak is also using JSP tags
> extensively.
> 3. my JSTL pages produce the memory leak every time.
> 4. static files and images do not produce the bug.
>
> One difference I can think of is JSTL accesses the pageContext
> frequently. My test pages use JSTL <c:import> quite a bit, which might
> be causing the leak? In either case, I sent my test pages to Remy.
After getting the 'string', I can get the memory problem with the given
example.
After looking into it, it happens also with the include and forward
examples, so it's either a bug in the page context, or more likely with
4.1's request dispatcher (since Jasper 2 + 4.0.x supposedly works well).
Given that Coyote + 4.0.x reportedly works fine, it is also unlikely there's
a problem there.
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by peter lin <pe...@labs.gte.com>.
Here is a bit more information on why I haven't ruled out JSTL or 1.2
JSP tags as a possible cause.
1. the standard examples all run fine and do not exhibit a memory leak,
including the simpletag example.
2. the other person experiencing memory leak is also using JSP tags
extensively.
3. my JSTL pages produce the memory leak every time.
4. static files and images do not produce the bug.
One difference I can think of is JSTL accesses the pageContext
frequently. My test pages use JSTL <c:import> quite a bit, which might
be causing the leak? In either case, I sent my test pages to Remy.
peter
Remy Maucherat wrote:
>
> > hey remy,
> >
> >
> > I'll send you more information in a hour with more details. I haven't
> > been able to find time to run the JProbe thread analyzer tests, but I
> > can atleast send you the test JSTL pages and JMeter settings I am using
> > to reproduce it.
> >
> > If that is cool with you, I'll get the files together and send it to
> > you.
>
> This is not very useful since you said the problem isn't with Jasper (JSTL
> is probably the least specific test you can come up with, more or less).
> If the problem really isn't with Jasper, I should be able to reproduce it
> with servlets and/or other JSPs, and I'm not.
>
> Remy
>
> --
> To unsubscribe, e-mail: <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
> I haven't ruled out JSTL as a cause since my test show it is memory
> intensive. I doubt it is JSTL alone, but it could be a combination.
> I've produced the bug with jdk1.3.1 on both windows and solaris. I sent
> you additional information and the exception I am getting. It's probably
> not going to be any help, but maybe it will rule out possible causes.
I can hammer the Transform example of JSTL with ab (concurrency 40) without
any memory problems (JDK 1.3 / Win2k). Other that there's one session
created per request, and that slowly eats up memory.
Your example uses another taglib. I'm going to try to download it, but I
really doubt there's a problem now.
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by peter lin <pe...@labs.gte.com>.
I haven't ruled out JSTL as a cause since my test show it is memory
intensive. I doubt it is JSTL alone, but it could be a combination.
I've produced the bug with jdk1.3.1 on both windows and solaris. I sent
you additional information and the exception I am getting. It's probably
not going to be any help, but maybe it will rule out possible causes.
peter
Remy Maucherat wrote:
>
> > hey remy,
> >
> >
> > I'll send you more information in a hour with more details. I haven't
> > been able to find time to run the JProbe thread analyzer tests, but I
> > can atleast send you the test JSTL pages and JMeter settings I am using
> > to reproduce it.
> >
> > If that is cool with you, I'll get the files together and send it to
> > you.
>
> This is not very useful since you said the problem isn't with Jasper (JSTL
> is probably the least specific test you can come up with, more or less).
> If the problem really isn't with Jasper, I should be able to reproduce it
> with servlets and/or other JSPs, and I'm not.
>
> Remy
>
> --
> To unsubscribe, e-mail: <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
> hey remy,
>
>
> I'll send you more information in a hour with more details. I haven't
> been able to find time to run the JProbe thread analyzer tests, but I
> can atleast send you the test JSTL pages and JMeter settings I am using
> to reproduce it.
>
> If that is cool with you, I'll get the files together and send it to
> you.
This is not very useful since you said the problem isn't with Jasper (JSTL
is probably the least specific test you can come up with, more or less).
If the problem really isn't with Jasper, I should be able to reproduce it
with servlets and/or other JSPs, and I'm not.
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by peter lin <pe...@labs.gte.com>.
hey remy,
I'll send you more information in a hour with more details. I haven't
been able to find time to run the JProbe thread analyzer tests, but I
can atleast send you the test JSTL pages and JMeter settings I am using
to reproduce it.
If that is cool with you, I'll get the files together and send it to
you.
peter
Remy Maucherat wrote:
>
> > I just noticed a memory leak in a build from CVS in the last 24 hours.
> > I am working on tracking it down.
>
> I have seen that kind of report on tomcat-user, but people have been unable
> to help reproduce it. They've tried with Coyote HTTP/1.1 + TC 4.0.x (no
> leak), as well as Jasper 2 + TC 4.0.x (no leak). It's very odd there's a
> memory leak in the Catalina core.
>
> I tried reproducing it using servlet and JSP based tests, but without any
> success.
>
> Remy
>
> --
> To unsubscribe, e-mail: <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Peter Lin <tc...@yahoo.com>.
hey remy,
I am running JProbe now. finally I have time to run
the test. If I find anything I'll post it.
peter
--- Remy Maucherat <re...@apache.org> wrote:
> Glenn Nielsen wrote:
> > I have been able to reproduce the memory leak with
> the following config:
> >
> > Apache 1.3.26/mod_jk1.2/Ajp13/Tomcat4.1 and
> Jasper1 built from CVS HEAD.
> > A JSP page which uses the dbtags taglib, a JNDI
> named JDBC DataSource,
> > and the DBCP as the datasource exhibits the
> problem.
> >
> > I am currently doing memory profiling using
> OptimizeIt to see if I can
> > find it.
>
> I have fixed the memory leak (and optimized the
> request dispatcher a bit
> in the process). I fail to understand why the memory
> was leaked, as I
> couldn't find any references kept anywhere (maybe it
> was a GC bug ?).
> Please confirm it is fixed if you can.
> The VM crash also went away after updating my CVS.
>
> So everything looks good, and I'll tag 4.1.6
> tomorrow.
>
> Remy
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Peter Lin <tc...@yahoo.com>.
after 3 hours of trying to get JProbe to run correctly
I give up. It's 12am and I'm too tired. If anyone
knows how to get tomcat 4.1.5, jasper2 and JProbe
profiler to work correctly I'll try to run the tests.
since jasper2 uses ANT and I don't see how to pass
Java_home from jprobe to the shell it won't compile
pages.
peter
--- Remy Maucherat <re...@apache.org> wrote:
> Glenn Nielsen wrote:
> > I have been able to reproduce the memory leak with
> the following config:
> >
> > Apache 1.3.26/mod_jk1.2/Ajp13/Tomcat4.1 and
> Jasper1 built from CVS HEAD.
> > A JSP page which uses the dbtags taglib, a JNDI
> named JDBC DataSource,
> > and the DBCP as the datasource exhibits the
> problem.
> >
> > I am currently doing memory profiling using
> OptimizeIt to see if I can
> > find it.
>
> I have fixed the memory leak (and optimized the
> request dispatcher a bit
> in the process). I fail to understand why the memory
> was leaked, as I
> couldn't find any references kept anywhere (maybe it
> was a GC bug ?).
> Please confirm it is fixed if you can.
> The VM crash also went away after updating my CVS.
>
> So everything looks good, and I'll tag 4.1.6
> tomorrow.
>
> Remy
>
>
> --
> To unsubscribe, e-mail:
> <ma...@jakarta.apache.org>
> For additional commands, e-mail:
> <ma...@jakarta.apache.org>
>
__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
Remy Maucherat wrote:
>> I think we were each tracking down a different memory leak.
>> The one I found is in DBCP, I should have that patched sometime tomorrow.
>
>
> Oh, ok.
> I'll wait a bit for your commit (and I'll apply the Tyrex 1.0 bug I fixed).
>
> It is definitely not as critical that the memory leak I fixed (using the
> RD was leaking memory, although it really looks like a JDK bug instead
> of a TC bug).
On second thought, I'll have trouble picking up your fix. Since 4.1.6
will be a beta at most (and given that a lot of stuff was added over
4.1.5, there may be problems), and the bug looks rerlatively minor
except in a demanding production evironment, I'll package it without
this fix.
4.1.6 will be focusing on testing:
- that the RD memory leak is fixed (that was the major issue)
- tag reuse in Jasper 2 (everything looks good so far, but few people
tried Jasper since 4.1.3, where it was found to be solid); the
possibility of disabling that has been added if non-compliant tags are
being used
I'll post an announcement about the test milestone to tc-user, as many
users have started testing 4.1.3, and had trouble with the memory leak,
and may want to give 4.1.x another shot. Sorry for taking so long for
finding/fixing it.
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
> I think we were each tracking down a different memory leak.
> The one I found is in DBCP, I should have that patched sometime tomorrow.
Oh, ok.
I'll wait a bit for your commit (and I'll apply the Tyrex 1.0 bug I fixed).
It is definitely not as critical that the memory leak I fixed (using the
RD was leaking memory, although it really looks like a JDK bug instead
of a TC bug).
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
Glenn Nielsen wrote:
> Glenn Nielsen wrote:
>
> I have tested and committed my patches to fix the memory leak in DBCP.
Cool.
Could you send me the new binary by email ? (that would be easier for me)
Thanks,
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Glenn Nielsen <gl...@voyager.apg.more.net>.
Glenn Nielsen wrote:
>
> Remy Maucherat wrote:
> >
> > Glenn Nielsen wrote:
> > > I have been able to reproduce the memory leak with the following config:
> > >
> > > Apache 1.3.26/mod_jk1.2/Ajp13/Tomcat4.1 and Jasper1 built from CVS HEAD.
> > > A JSP page which uses the dbtags taglib, a JNDI named JDBC DataSource,
> > > and the DBCP as the datasource exhibits the problem.
> > >
> > > I am currently doing memory profiling using OptimizeIt to see if I can
> > > find it.
> >
> > I have fixed the memory leak (and optimized the request dispatcher a bit
> > in the process). I fail to understand why the memory was leaked, as I
> > couldn't find any references kept anywhere (maybe it was a GC bug ?).
> > Please confirm it is fixed if you can.
> > The VM crash also went away after updating my CVS.
> >
> > So everything looks good, and I'll tag 4.1.6 tomorrow.
> >
> > Remy
> >
>
> I think we were each tracking down a different memory leak.
> The one I found is in DBCP, I should have that patched sometime tomorrow.
>
I have tested and committed my patches to fix the memory leak in DBCP.
Regards,
Glenn
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Glenn Nielsen <gl...@voyager.apg.more.net>.
Remy Maucherat wrote:
>
> Glenn Nielsen wrote:
> > I have been able to reproduce the memory leak with the following config:
> >
> > Apache 1.3.26/mod_jk1.2/Ajp13/Tomcat4.1 and Jasper1 built from CVS HEAD.
> > A JSP page which uses the dbtags taglib, a JNDI named JDBC DataSource,
> > and the DBCP as the datasource exhibits the problem.
> >
> > I am currently doing memory profiling using OptimizeIt to see if I can
> > find it.
>
> I have fixed the memory leak (and optimized the request dispatcher a bit
> in the process). I fail to understand why the memory was leaked, as I
> couldn't find any references kept anywhere (maybe it was a GC bug ?).
> Please confirm it is fixed if you can.
> The VM crash also went away after updating my CVS.
>
> So everything looks good, and I'll tag 4.1.6 tomorrow.
>
> Remy
>
I think we were each tracking down a different memory leak.
The one I found is in DBCP, I should have that patched sometime tomorrow.
Regards,
Glenn
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
Glenn Nielsen wrote:
> I have been able to reproduce the memory leak with the following config:
>
> Apache 1.3.26/mod_jk1.2/Ajp13/Tomcat4.1 and Jasper1 built from CVS HEAD.
> A JSP page which uses the dbtags taglib, a JNDI named JDBC DataSource,
> and the DBCP as the datasource exhibits the problem.
>
> I am currently doing memory profiling using OptimizeIt to see if I can
> find it.
I have fixed the memory leak (and optimized the request dispatcher a bit
in the process). I fail to understand why the memory was leaked, as I
couldn't find any references kept anywhere (maybe it was a GC bug ?).
Please confirm it is fixed if you can.
The VM crash also went away after updating my CVS.
So everything looks good, and I'll tag 4.1.6 tomorrow.
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Glenn Nielsen <gl...@voyager.apg.more.net>.
I have been able to reproduce the memory leak with the following config:
Apache 1.3.26/mod_jk1.2/Ajp13/Tomcat4.1 and Jasper1 built from CVS HEAD.
A JSP page which uses the dbtags taglib, a JNDI named JDBC DataSource,
and the DBCP as the datasource exhibits the problem.
I am currently doing memory profiling using OptimizeIt to see if I can
find it.
Regards,
Glenn
Remy Maucherat wrote:
>
> > I just noticed a memory leak in a build from CVS in the last 24 hours.
> > I am working on tracking it down.
>
> I have seen that kind of report on tomcat-user, but people have been unable
> to help reproduce it. They've tried with Coyote HTTP/1.1 + TC 4.0.x (no
> leak), as well as Jasper 2 + TC 4.0.x (no leak). It's very odd there's a
> memory leak in the Catalina core.
>
> I tried reproducing it using servlet and JSP based tests, but without any
> success.
>
> Remy
>
> --
> To unsubscribe, e-mail: <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
> I just noticed a memory leak in a build from CVS in the last 24 hours.
> I am working on tracking it down.
I have seen that kind of report on tomcat-user, but people have been unable
to help reproduce it. They've tried with Coyote HTTP/1.1 + TC 4.0.x (no
leak), as well as Jasper 2 + TC 4.0.x (no leak). It's very odd there's a
memory leak in the Catalina core.
I tried reproducing it using servlet and JSP based tests, but without any
success.
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Glenn Nielsen <gl...@voyager.apg.more.net>.
I just noticed a memory leak in a build from CVS in the last 24 hours.
I am working on tracking it down.
Glenn
Remy Maucherat wrote:
>
> > > There are only a few issues remaining:
> > > - Updating JNDI resources with the admin webapp is not dynamic (for
> > > reasons currently beyond my understanding). Doing a stop/start on the
> > > context allows to pick up the changes, so the bug is only minor.
> > > - Nacho's IIS issues with JK 2.
> > > - Costin's bug with Jasper 2.
>
> Since the last two issues are resolved, I plan to tag the new 4.1.6
> milestone later today.
>
> Remy
>
> --
> To unsubscribe, e-mail: <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by Remy Maucherat <re...@apache.org>.
> > There are only a few issues remaining:
> > - Updating JNDI resources with the admin webapp is not dynamic (for
> > reasons currently beyond my understanding). Doing a stop/start on the
> > context allows to pick up the changes, so the bug is only minor.
> > - Nacho's IIS issues with JK 2.
> > - Costin's bug with Jasper 2.
Since the last two issues are resolved, I plan to tag the new 4.1.6
milestone later today.
Remy
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>
Re: [4.1.6] New milestone release soon
Posted by co...@covalent.net.
On Wed, 26 Jun 2002, Remy Maucherat wrote:
> There are only a few issues remaining:
> - Updating JNDI resources with the admin webapp is not dynamic (for
> reasons currently beyond my understanding). Doing a stop/start on the
> context allows to pick up the changes, so the bug is only minor.
> - Nacho's IIS issues with JK 2.
> - Costin's bug with Jasper 2.
I'll verify again ( I'm curently using 4.1.2 snapshot of jasper2 ).
I think the bug was pretty much a showstopper - we're talking about
trivial and common uses of struts ( and I assume jstl or any tag libs
that have iterations and conditions ).
Costin
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>