> Memory will never be released if you have strong references to the

> ensure that objects that use unmanaged
> resources have implemented Dispose method.

> Try to use CLR Profiler from Microsoft, it will show you the heap map &
the
> amount of memory objects of particular type consume....

"Vadym Stetsyak" <pdsdev@ukr.net> wrote in message
news:e2vGqoz5DHA.2736@TK2MSFTNGP09.phx.gbl...
> Memory will never be released if you have strong references to the
objects.
> Try to use weak references, or ensure that objects that use unmanaged
> resources have implemented Dispose method.
>
> Try to use CLR Profiler from Microsoft, it will show you the heap map &
the
> amount of memory objects of particular type consume....
>
> "Nick Carter" <nickcarter@goawayspam.com> wrote in message
> news:eoNewgz5DHA.1948@TK2MSFTNGP12.phx.gbl...
> > I am searching for a suspected memory leak and I have made some
> assumptions
> > and I would be
> > grateful if someone could check my thinking...
> >
> > I have a remoting server that I believe has a memory problem. My first
> > assumption is that I
> > believe it has a memory problem because I am checking "Mem Usage" in
Task
> > Manager. One
> > morning the Mem Usage is, say, 100Mb and the next morning it is, say,
> 200Mb
> > (after a
> > complete enforced garbage collection) and the next morning it is higher
> yet
> > again.
> > Eventually the remoting server throws an out of memory exception. So my
> > first assumption is
> > that it is safe to use the Mem Usage figure in this way. I know I could
> use
> > .NET's memory
> > performance counters instead but Task Manager was certainly present on
the
> > target machine
> > so it was quick and easy.
> >
> > My second assumption is that the increase in memory is a consequence of
a
> > memory leak. I
> > wonder if this is actually true. I am wondering if the garbage collector
> is
> > resizing
> > generation 0, 1 and 2 so that they are bigger so that they can more
easily
> > cope with the
> > load that it is under. If this were the case then surely the increase in
> > memory requirement
> > is simply normal behaviour and not a sign of a memory leak.
> >
> > I am using SciTech's .NET Memory Profiler to see if there is a memory
> leak.
> > I am new to this
> > tool so I could be using it wrong. I take a snapshot once the remoting
> > server is up and
> > running and has been doing a little bit of normal processing. Then when
I
> > think the memory
> > requirement has gone up unnecessarily and is not coming down again (even
> > after a garbage
> > collection) I take another snapshot and compare them. My next assumption
> is
> > that if the
> > "Delta" column shows "0" differences for all classes between the two
> > snapshots then there is
> > no memory leak. Is this right ?
> >
> > I am also wondering if the memory leak could be in unmanaged code
instead
> of
> > managed code. I
> > am not familiar with how unmanaged code handles memory and if I would
see
> a
> > memory leak that
> > originated from unmanaged code.
> >
> > Any pointers on any of these issues would be gratefully received.
> >
> > Thanks.
> >
> > Nick Carter
> >
> >
>
>

"Nick Carter" <nickcarter@goawayspam.com> wrote in message
news:eoNewgz5DHA.1948@TK2MSFTNGP12.phx.gbl...
> I am searching for a suspected memory leak and I have made some
assumptions
> and I would be
> grateful if someone could check my thinking...
>
> I have a remoting server that I believe has a memory problem. My first
> assumption is that I
> believe it has a memory problem because I am checking "Mem Usage" in Task
> Manager. One
> morning the Mem Usage is, say, 100Mb and the next morning it is, say,
200Mb
> (after a
> complete enforced garbage collection) and the next morning it is higher
yet
> again.
> Eventually the remoting server throws an out of memory exception. So my
> first assumption is
> that it is safe to use the Mem Usage figure in this way. I know I could
use
> .NET's memory
> performance counters instead but Task Manager was certainly present on the
> target machine
> so it was quick and easy.
>
> My second assumption is that the increase in memory is a consequence of a
> memory leak. I
> wonder if this is actually true. I am wondering if the garbage collector
is
> resizing
> generation 0, 1 and 2 so that they are bigger so that they can more easily
> cope with the
> load that it is under. If this were the case then surely the increase in
> memory requirement
> is simply normal behaviour and not a sign of a memory leak.
>
> I am using SciTech's .NET Memory Profiler to see if there is a memory
leak.
> I am new to this
> tool so I could be using it wrong. I take a snapshot once the remoting
> server is up and
> running and has been doing a little bit of normal processing. Then when I
> think the memory
> requirement has gone up unnecessarily and is not coming down again (even
> after a garbage
> collection) I take another snapshot and compare them. My next assumption
is
> that if the
> "Delta" column shows "0" differences for all classes between the two
> snapshots then there is
> no memory leak. Is this right ?
>
> I am also wondering if the memory leak could be in unmanaged code instead
of
> managed code. I
> am not familiar with how unmanaged code handles memory and if I would see
a
> memory leak that
> originated from unmanaged code.
>
> Any pointers on any of these issues would be gratefully received.
>
> Thanks.
>
> Nick Carter
>
>

"Nick Carter" <nickcarter@goawayspam.com> wrote in message
news:usjEk3z5DHA.2748@TK2MSFTNGP09.phx.gbl...
> Vadym,
>
> > Memory will never be released if you have strong references to the
> objects.
>
> The references to objects are released (because the references go out of
> scope) so the objects are released.
>
> > ensure that objects that use unmanaged
> > resources have implemented Dispose method.
>
> Fortunately we have no objects which use unmanaged resources (except for
the
> classes in the .NET Framework which uses unmanaged resources).
>
> > Try to use CLR Profiler from Microsoft, it will show you the heap map &
> the
> > amount of memory objects of particular type consume....
>
> Thanks. I started to look at it but it was a bit daunting. I will take
> another look.
> Nick
>
> "Vadym Stetsyak" <pdsdev@ukr.net> wrote in message
> news:e2vGqoz5DHA.2736@TK2MSFTNGP09.phx.gbl...
> > Memory will never be released if you have strong references to the
> objects.
> > Try to use weak references, or ensure that objects that use unmanaged
> > resources have implemented Dispose method.
> >
> > Try to use CLR Profiler from Microsoft, it will show you the heap map &
> the
> > amount of memory objects of particular type consume....
> >
> > "Nick Carter" <nickcarter@goawayspam.com> wrote in message
> > news:eoNewgz5DHA.1948@TK2MSFTNGP12.phx.gbl...
> > > I am searching for a suspected memory leak and I have made some
> > assumptions
> > > and I would be
> > > grateful if someone could check my thinking...
> > >
> > > I have a remoting server that I believe has a memory problem. My first
> > > assumption is that I
> > > believe it has a memory problem because I am checking "Mem Usage" in
> Task
> > > Manager. One
> > > morning the Mem Usage is, say, 100Mb and the next morning it is, say,
> > 200Mb
> > > (after a
> > > complete enforced garbage collection) and the next morning it is
higher
> > yet
> > > again.
> > > Eventually the remoting server throws an out of memory exception. So
my
> > > first assumption is
> > > that it is safe to use the Mem Usage figure in this way. I know I
could
> > use
> > > .NET's memory
> > > performance counters instead but Task Manager was certainly present on
> the
> > > target machine
> > > so it was quick and easy.
> > >
> > > My second assumption is that the increase in memory is a consequence
of
> a
> > > memory leak. I
> > > wonder if this is actually true. I am wondering if the garbage
collector
> > is
> > > resizing
> > > generation 0, 1 and 2 so that they are bigger so that they can more
> easily
> > > cope with the
> > > load that it is under. If this were the case then surely the increase
in
> > > memory requirement
> > > is simply normal behaviour and not a sign of a memory leak.
> > >
> > > I am using SciTech's .NET Memory Profiler to see if there is a memory
> > leak.
> > > I am new to this
> > > tool so I could be using it wrong. I take a snapshot once the remoting
> > > server is up and
> > > running and has been doing a little bit of normal processing. Then
when
> I
> > > think the memory
> > > requirement has gone up unnecessarily and is not coming down again
(even
> > > after a garbage
> > > collection) I take another snapshot and compare them. My next
assumption
> > is
> > > that if the
> > > "Delta" column shows "0" differences for all classes between the two
> > > snapshots then there is
> > > no memory leak. Is this right ?
> > >
> > > I am also wondering if the memory leak could be in unmanaged code
> instead
> > of
> > > managed code. I
> > > am not familiar with how unmanaged code handles memory and if I would
> see
> > a
> > > memory leak that
> > > originated from unmanaged code.
> > >
> > > Any pointers on any of these issues would be gratefully received.
> > >
> > > Thanks.
> > >
> > > Nick Carter
> > >
> > >
> >
> >
>
>

"Niall" <asdf@me.com> wrote in message
news:u5QQqcr6DHA.1664@TK2MSFTNGP11.phx.gbl...
> Nick,
>
> I have been using the .Net Memory Profiler pretty heavily for quite some
> time. At first, it can be a little confusing, but I have not found another
> tool that gives you the information as usefully as this profiler. The CLR
> Profiler is good, but if your app is of a decent size, you can sit there
all
> day watching it draw pretty pictures.
>
> What the .Net Memory Profiler provides which I find most useful is the
list
> of root paths to an object. As far as I can see, the CLR Profiler and all
> other memory profilers I've tried just show you all references, which is a
> lot when you are looking at windows forms (event handlers are everywhere).
> The root paths are the reasons why the object isn't collectible, and hence
> the root paths are the only information you need to find a memory leak.
>
> So. The way you are using the profiler sounds correct to me. Take a
> snapshot, let it do some work and get to a stage where you think its
memory
> should be freed, and take another snapshot. Go View->Class Filter->Show
> Classes With New Or Removed Instances. This will basically show you what
was
> created or collected between the two snapshots.
>
> You can then look through the list of classes and find objects that you
> think should have been created and then collected, but the profiler shows
as
> still alive. Double click on the suspicious classes to go to the list of
> instances. You can have a toy with the allocation stacks to find which
> instances you're interested in, then pick one and double click it to see
its
> details. The list of root paths that show are the reasons why the object
has
> not been collected. From this, you can trace it back to the the source of
> the problem and change your code to free up the appropriate references.
>
> Be aware that the profiler does a full GC when taking a snapshot. Hope
this
> helps!
>
> Niall
>
> "Nick Carter" <nickcarter@goawayspam.com> wrote in message
> news:usjEk3z5DHA.2748@TK2MSFTNGP09.phx.gbl...
> > Vadym,
> >
> > > Memory will never be released if you have strong references to the
> > objects.
> >
> > The references to objects are released (because the references go out of
> > scope) so the objects are released.
> >
> > > ensure that objects that use unmanaged
> > > resources have implemented Dispose method.
> >
> > Fortunately we have no objects which use unmanaged resources (except for
> the
> > classes in the .NET Framework which uses unmanaged resources).
> >
> > > Try to use CLR Profiler from Microsoft, it will show you the heap map
&
> > the
> > > amount of memory objects of particular type consume....
> >
> > Thanks. I started to look at it but it was a bit daunting. I will take
> > another look.
> > Nick
> >
> > "Vadym Stetsyak" <pdsdev@ukr.net> wrote in message
> > news:e2vGqoz5DHA.2736@TK2MSFTNGP09.phx.gbl...
> > > Memory will never be released if you have strong references to the
> > objects.
> > > Try to use weak references, or ensure that objects that use unmanaged
> > > resources have implemented Dispose method.
> > >
> > > Try to use CLR Profiler from Microsoft, it will show you the heap map
&
> > the
> > > amount of memory objects of particular type consume....
> > >
> > > "Nick Carter" <nickcarter@goawayspam.com> wrote in message
> > > news:eoNewgz5DHA.1948@TK2MSFTNGP12.phx.gbl...
> > > > I am searching for a suspected memory leak and I have made some
> > > assumptions
> > > > and I would be
> > > > grateful if someone could check my thinking...
> > > >
> > > > I have a remoting server that I believe has a memory problem. My
first
> > > > assumption is that I
> > > > believe it has a memory problem because I am checking "Mem Usage" in
> > Task
> > > > Manager. One
> > > > morning the Mem Usage is, say, 100Mb and the next morning it is,
say,
> > > 200Mb
> > > > (after a
> > > > complete enforced garbage collection) and the next morning it is
> higher
> > > yet
> > > > again.
> > > > Eventually the remoting server throws an out of memory exception. So
> my
> > > > first assumption is
> > > > that it is safe to use the Mem Usage figure in this way. I know I
> could
> > > use
> > > > .NET's memory
> > > > performance counters instead but Task Manager was certainly present
on
> > the
> > > > target machine
> > > > so it was quick and easy.
> > > >
> > > > My second assumption is that the increase in memory is a consequence
> of
> > a
> > > > memory leak. I
> > > > wonder if this is actually true. I am wondering if the garbage
> collector
> > > is
> > > > resizing
> > > > generation 0, 1 and 2 so that they are bigger so that they can more
> > easily
> > > > cope with the
> > > > load that it is under. If this were the case then surely the
increase
> in
> > > > memory requirement
> > > > is simply normal behaviour and not a sign of a memory leak.
> > > >
> > > > I am using SciTech's .NET Memory Profiler to see if there is a
memory
> > > leak.
> > > > I am new to this
> > > > tool so I could be using it wrong. I take a snapshot once the
remoting
> > > > server is up and
> > > > running and has been doing a little bit of normal processing. Then
> when
> > I
> > > > think the memory
> > > > requirement has gone up unnecessarily and is not coming down again
> (even
> > > > after a garbage
> > > > collection) I take another snapshot and compare them. My next
> assumption
> > > is
> > > > that if the
> > > > "Delta" column shows "0" differences for all classes between the two
> > > > snapshots then there is
> > > > no memory leak. Is this right ?
> > > >
> > > > I am also wondering if the memory leak could be in unmanaged code
> > instead
> > > of
> > > > managed code. I
> > > > am not familiar with how unmanaged code handles memory and if I
would
> > see
> > > a
> > > > memory leak that
> > > > originated from unmanaged code.
> > > >
> > > > Any pointers on any of these issues would be gratefully received.
> > > >
> > > > Thanks.
> > > >
> > > > Nick Carter
> > > >
> > > >
> > >
> > >
> >
> >
>
>