wrapper-user

[Wrapper-user] Wrapper and Time

[Grant \(ProtectionNET\) <gr...@pr...>]

Hi,

I have a case where the Java app changes the system time (via OS calls =
etc) and when Wrapper is running, I have sometimes seen strange problems =
and the wrapper service shuts down or goes crazy when the time is =
changed around.

I'm only starting to investigate this now, so I don't have all the facts =
yet.

Can someone explain to me what reliance Wrapper has to time and how it =
uses it... that would help me track this problem down...=20

Thanks a billion.... and merry-xmas and all that ;-)

Regards,

Grant 

Re: [Wrapper-user] Wrapper and Time

[Leif M. <le...@ta...>]

Grant,
    :-/  That doesn't surprise me in the least that you are having 
problems there.  The
Wrapper makes extensive use of time both in its native and Java 
components to keep
track of various timeouts.   The Wrapper currently relies on the system 
time being
fairly reliable.

    I could probably add some code to reset the various timeouts in the 
event that either
component of the Wrapper detects that the system time appears to have 
been changed.
Ie when the time decreases or is increased by a large amount.

    Small forward adjustments of the system time would be very difficult 
to detect
because the Wrapper contains features to detect when the system has been 
heavily
loaded.  I am not sure that it would be possible to differentiate 
between the cases where
the process is not given any CPU for a minute vs the clock being 
instantly set ahead
by a minute.

    Could you describe how you are adjusting the time?  Ie by how much, 
in which
direction and why?  It will help me to prioritize this.   If you know 
that the Wrapper
simply does not like having the clock changed, would it be unreasonable 
to just not
do this? Or is changing the clock a common operation?

    I know I can make the Wrapper behave better than it is now, but am 
unsure of how
perfect I can make it perform under all situations.

Cheers,
Leif

Grant (ProtectionNET) wrote:

> Hi,
>  
> I have a case where the Java app changes the system time (via OS calls 
> etc) and when Wrapper is running, I have sometimes seen strange 
> problems and the wrapper service shuts down or goes crazy when the 
> time is changed around.
>  
> I'm only starting to investigate this now, so I don't have all the 
> facts yet.
>  
> Can someone explain to me what reliance Wrapper has to time and how it 
> uses it... that would help me track this problem down...
>  
> Thanks a billion.... and merry-xmas and all that ;-)
>  
> Regards,
>  
> Grant

Re: [Wrapper-user] Wrapper and Time

[Grant \(ProtectionNET\) <gr...@pr...>]

Lief,

Changing the time and timezone is part of the system configuration that this
java app performs. Normally this would only occur at initial setup and
possibly only rarely so I'm going to avoid the whole issue by making the app
restart. I'm obviously still caught in a situation where wrapper still
thinks time has gone silly...

It seems as though setting the time backwards (ie changing it from 2:00pm to
10:00) causes Wrapper to report silly things like "process has not recieved
CPU time for 14308 seconds, extending timeouts" which worry me because what
is it doing when it see's this. I have also had the wrapper complain and
restart - not good.

I've not looked through the source of wrapper (perhaps I should) but why
could you not just increment an int for X times and sleep for 1 sec each
time. When you get to X you would know that the code has reached that
timeout - this makes your counters work regardless of what the real time is
outside.

I'm going to look through the source and see if I can make what I need to
work. If I figure anything out, I'll let you know.

Grant

----- Original Message ----- 
From: "Leif Mortenson" <le...@ta...>
To: <wra...@li...>
Sent: Friday, December 26, 2003 10:33 AM
Subject: Re: [Wrapper-user] Wrapper and Time


> Grant,
>     :-/  That doesn't surprise me in the least that you are having
> problems there.  The
> Wrapper makes extensive use of time both in its native and Java
> components to keep
> track of various timeouts.   The Wrapper currently relies on the system
> time being
> fairly reliable.
>
>     I could probably add some code to reset the various timeouts in the
> event that either
> component of the Wrapper detects that the system time appears to have
> been changed.
> Ie when the time decreases or is increased by a large amount.
>
>     Small forward adjustments of the system time would be very difficult
> to detect
> because the Wrapper contains features to detect when the system has been
> heavily
> loaded.  I am not sure that it would be possible to differentiate
> between the cases where
> the process is not given any CPU for a minute vs the clock being
> instantly set ahead
> by a minute.
>
>     Could you describe how you are adjusting the time?  Ie by how much,
> in which
> direction and why?  It will help me to prioritize this.   If you know
> that the Wrapper
> simply does not like having the clock changed, would it be unreasonable
> to just not
> do this? Or is changing the clock a common operation?
>
>     I know I can make the Wrapper behave better than it is now, but am
> unsure of how
> perfect I can make it perform under all situations.
>
> Cheers,
> Leif
>
> Grant (ProtectionNET) wrote:
>
> > Hi,
> >
> > I have a case where the Java app changes the system time (via OS calls
> > etc) and when Wrapper is running, I have sometimes seen strange
> > problems and the wrapper service shuts down or goes crazy when the
> > time is changed around.
> >
> > I'm only starting to investigate this now, so I don't have all the
> > facts yet.
> >
> > Can someone explain to me what reliance Wrapper has to time and how it
> > uses it... that would help me track this problem down...
> >
> > Thanks a billion.... and merry-xmas and all that ;-)
> >
> > Regards,
> >
> > Grant
>
>
>
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by: IBM Linux Tutorials.
> Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
> Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
> Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
> _______________________________________________
> Wrapper-user mailing list
> Wra...@li...
> https://lists.sourceforge.net/lists/listinfo/wrapper-user

Re: [Wrapper-user] Wrapper and Time

[Grant \(ProtectionNET\) <gr...@pr...>]

Leif,

Another thing, when DST comes into effect, some regions change +- by 2 hours
so this is going to cause a problem relying on time in the wrapper code.

I poked around in the wrapper.c and WrapperManager.java source and I
absolutely think you could modify the code to include a no-time dependant
counter.

This would require you to create your own timer thread that polls every
second and increments an unsigned long.

I use this code in some of my Java apps to get a counter that continues
forward independant of the actual system time.

private long secondTimer = 0;

Thread t = new Thread()
{
 public void run()
 {
  while (true)
  {
   try
   {
    secondTimer++;
    if (secondTimer == Long.MAX_VALUE-1) secondTimer = 0;
    sleep(1000);
   }
   catch (Exception e)
   {
   }
  }
 }

};
t.start();

I'm sure it wouldn't take long to develop the same code for the wrapper.c on
Win32 and Unix ? The lack of standard threads in unix could be an issue? but
using time the way you have will definitely cause an issue every DST
adjustment or when the user changes the system time. They'll need to restart
wrapper to fix this.

I would continue to use a date based timer for understanding CPU load but
ignore anything that jumps by more than 10 seconds as time adjustment above
this is likely a time change. 10 seconds with no CPU time is a LONG time. If
everything else moves to a CPU timer (on unix you could just count every 1/4
second or adjust that sleep to 1 second and continue counting).

When using NTP, the systems tend to drift time upwards and downwards
gracefully so that apps don't notice the change - most hwclocks will only
loose 1 second a day at worst so this would not really be worth worrying
about.

Grant

----- Original Message ----- 
From: "Leif Mortenson" <le...@ta...>
To: <wra...@li...>
Sent: Friday, December 26, 2003 10:33 AM
Subject: Re: [Wrapper-user] Wrapper and Time


> Grant,
>     :-/  That doesn't surprise me in the least that you are having
> problems there.  The
> Wrapper makes extensive use of time both in its native and Java
> components to keep
> track of various timeouts.   The Wrapper currently relies on the system
> time being
> fairly reliable.
>
>     I could probably add some code to reset the various timeouts in the
> event that either
> component of the Wrapper detects that the system time appears to have
> been changed.
> Ie when the time decreases or is increased by a large amount.
>
>     Small forward adjustments of the system time would be very difficult
> to detect
> because the Wrapper contains features to detect when the system has been
> heavily
> loaded.  I am not sure that it would be possible to differentiate
> between the cases where
> the process is not given any CPU for a minute vs the clock being
> instantly set ahead
> by a minute.
>
>     Could you describe how you are adjusting the time?  Ie by how much,
> in which
> direction and why?  It will help me to prioritize this.   If you know
> that the Wrapper
> simply does not like having the clock changed, would it be unreasonable
> to just not
> do this? Or is changing the clock a common operation?
>
>     I know I can make the Wrapper behave better than it is now, but am
> unsure of how
> perfect I can make it perform under all situations.
>
> Cheers,
> Leif
>
> Grant (ProtectionNET) wrote:
>
> > Hi,
> >
> > I have a case where the Java app changes the system time (via OS calls
> > etc) and when Wrapper is running, I have sometimes seen strange
> > problems and the wrapper service shuts down or goes crazy when the
> > time is changed around.
> >
> > I'm only starting to investigate this now, so I don't have all the
> > facts yet.
> >
> > Can someone explain to me what reliance Wrapper has to time and how it
> > uses it... that would help me track this problem down...
> >
> > Thanks a billion.... and merry-xmas and all that ;-)
> >
> > Regards,
> >
> > Grant
>
>
>
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by: IBM Linux Tutorials.
> Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
> Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
> Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
> _______________________________________________
> Wrapper-user mailing list
> Wra...@li...
> https://lists.sourceforge.net/lists/listinfo/wrapper-user

Re: [Wrapper-user] Wrapper and Time

[Leif M. <le...@ta...>]

Grant,
    Thanks for pointing this out.  The Wrapper has always worked great 
in an environment
where the system time is reliable but I agree that the changes you 
suggest would be a
good idea.  The changes are easy on the Java side, let me look into ways 
of making things
work in the native code.  I'll see if I can get this into the next 
release.  I'll post back if I
run into any problems.

Cheers,
Leif

Grant (ProtectionNET) wrote:

>Leif,
>
>Another thing, when DST comes into effect, some regions change +- by 2 hours
>so this is going to cause a problem relying on time in the wrapper code.
>
>I poked around in the wrapper.c and WrapperManager.java source and I
>absolutely think you could modify the code to include a no-time dependant
>counter.
>
>This would require you to create your own timer thread that polls every
>second and increments an unsigned long.
>
>I use this code in some of my Java apps to get a counter that continues
>forward independant of the actual system time.
>
>private long secondTimer = 0;
>
>Thread t = new Thread()
>{
> public void run()
> {
>  while (true)
>  {
>   try
>   {
>    secondTimer++;
>    if (secondTimer == Long.MAX_VALUE-1) secondTimer = 0;
>    sleep(1000);
>   }
>   catch (Exception e)
>   {
>   }
>  }
> }
>
>};
>t.start();
>
>I'm sure it wouldn't take long to develop the same code for the wrapper.c on
>Win32 and Unix ? The lack of standard threads in unix could be an issue? but
>using time the way you have will definitely cause an issue every DST
>adjustment or when the user changes the system time. They'll need to restart
>wrapper to fix this.
>
>I would continue to use a date based timer for understanding CPU load but
>ignore anything that jumps by more than 10 seconds as time adjustment above
>this is likely a time change. 10 seconds with no CPU time is a LONG time. If
>everything else moves to a CPU timer (on unix you could just count every 1/4
>second or adjust that sleep to 1 second and continue counting).
>
>When using NTP, the systems tend to drift time upwards and downwards
>gracefully so that apps don't notice the change - most hwclocks will only
>loose 1 second a day at worst so this would not really be worth worrying
>about.
>
>Grant
>
>----- Original Message ----- 
>From: "Leif Mortenson" <le...@ta...>
>To: <wra...@li...>
>Sent: Friday, December 26, 2003 10:33 AM
>Subject: Re: [Wrapper-user] Wrapper and Time
>
>
>  
>
>>Grant,
>>    :-/  That doesn't surprise me in the least that you are having
>>problems there.  The
>>Wrapper makes extensive use of time both in its native and Java
>>components to keep
>>track of various timeouts.   The Wrapper currently relies on the system
>>time being
>>fairly reliable.
>>
>>    I could probably add some code to reset the various timeouts in the
>>event that either
>>component of the Wrapper detects that the system time appears to have
>>been changed.
>>Ie when the time decreases or is increased by a large amount.
>>
>>    Small forward adjustments of the system time would be very difficult
>>to detect
>>because the Wrapper contains features to detect when the system has been
>>heavily
>>loaded.  I am not sure that it would be possible to differentiate
>>between the cases where
>>the process is not given any CPU for a minute vs the clock being
>>instantly set ahead
>>by a minute.
>>
>>    Could you describe how you are adjusting the time?  Ie by how much,
>>in which
>>direction and why?  It will help me to prioritize this.   If you know
>>that the Wrapper
>>simply does not like having the clock changed, would it be unreasonable
>>to just not
>>do this? Or is changing the clock a common operation?
>>
>>    I know I can make the Wrapper behave better than it is now, but am
>>unsure of how
>>perfect I can make it perform under all situations.
>>
>>Cheers,
>>Leif
>>
>>Grant (ProtectionNET) wrote:
>>
>>    
>>
>>>Hi,
>>>
>>>I have a case where the Java app changes the system time (via OS calls
>>>etc) and when Wrapper is running, I have sometimes seen strange
>>>problems and the wrapper service shuts down or goes crazy when the
>>>time is changed around.
>>>
>>>I'm only starting to investigate this now, so I don't have all the
>>>facts yet.
>>>
>>>Can someone explain to me what reliance Wrapper has to time and how it
>>>uses it... that would help me track this problem down...
>>>
>>>Thanks a billion.... and merry-xmas and all that ;-)
>>>
>>>Regards,
>>>
>>>Grant
>>>      
>>>
>>
>>
>>
>>-------------------------------------------------------
>>This SF.net email is sponsored by: IBM Linux Tutorials.
>>Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
>>Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
>>Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
>>_______________________________________________
>>Wrapper-user mailing list
>>Wra...@li...
>>https://lists.sourceforge.net/lists/listinfo/wrapper-user
>>    
>>
>
>
>
>-------------------------------------------------------
>This SF.net email is sponsored by: IBM Linux Tutorials.
>Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
>Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
>Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
>_______________________________________________
>Wrapper-user mailing list
>Wra...@li...
>https://lists.sourceforge.net/lists/listinfo/wrapper-user
>
>  
>