Re: [Wrapper-user] disabling automatic shutdowns

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

Shmul,
    Ok.  There were actually three bugs in the Wrapper related to this 
issue.  They
have all been fixed for the 3.1.0 release.

1) Spaces around the '=' character in a property definition were not 
being trimmed
when the wrapper.conf file was being parsed.  This was causing the 
property name
to appear to be "wrapper.ping.timeout " rather than 
"wrapper.ping.timeout", thus
causing the Wrapper to ignore your settings.

2) The very first ping timeout set after the JVM was started was still 
hard coded at
30 seconds.  This was causing the Wrapper to timeout after 30 seconds 
even though
the ping timeout was disabled.  If at least one ping had been received 
then things
would have worked correctly.   Not sure how this made it past me.

3) This one is minor.  But when the ping timeout is extended, the the 
SO_TIMEOUT
set on the socket between the Wrapper and its JVM also needs to be 
extended.  This
was happening for large values of wrapper.ping.timeout, but not for a 
value of 0.
This was leading to the "Read Timed Out" messages that you were seeing 
in the debug
log output.

As for what you can do now...

Since that first timeout is hard coded in 3.0.5, it is not possible to 
set the
wrapper.ping.interval greater that 30 seconds.  I would set it to 25 
seconds.
Then set the wrapper.ping.timeout to 0.   This will get rid of 4/5ths of 
the ping
messages from your wrapper.log and will also make sure that the Wrapper
never restarts the JVM even after it stops responding to pings.

Let me know how this works for you.

Cheers,
Leif


Shmulik Regev wrote:

>On 15 Mar 2004 at 2:16, Leif Mortenson wrote:
>
>  
>
>>    I was able to reproduce this using those timeout values.  It is
>>    late 
>>here, so it is not
>>immediately obvious what the problem is.  I will take a look at it. 
>>    
>>
>
>thanks.
>
>  
>
>>But for now please go back and remove the three timeout / interval
>>properties below.  What was the original reason that you had wanted to
>>set them?   Setting the timeout to 0 works. But the long ping interval
>>seems to be causing problems right now...
>>    
>>
>We originally started messing with these properties after we noticed 
>the jvm is restarted by the wrapper. this is probably due to a bug in 
>our application (that causes extensive cpU usage) which is hard to 
>debug when the wrapper restarts the application below our feet.
>
>Cheers,
>Shmul 
>  
>