You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@tomcat.apache.org by Arshiya Shariff <ar...@ericsson.com.INVALID> on 2020/09/21 16:56:17 UTC

HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Hi All,

The client has configured a response timeout of 1.5 seconds. In a case when our application tries to respond over a http2 stream asynchronously after 2 seconds where the client has already sent RST_STREAM with NO ERROR in 1.5 seconds (due to no response) , then tomcat sends GOAWAY and closes the HTTP2 connection . Is this behavior of GOAWAY and connection closure expected ?
We have planned to upgrade to Embedded tomcat version 9.0.38 . These are the behaviors we see in production with version 9.0.22 ,  so we need some help with analyzing / validating  the existing behavior before the upgrade .
Please let us know.

Thanks and Regards
Arshiya Shariff

Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Posted by Mark Thomas <ma...@apache.org>.
On 25/09/2020 15:47, Arshiya Shariff wrote:
> Thanks for the response Martin.
> 
> Below is the only exception we are getting in the logs while sending response to a closed stream , few times with MUST_COMPLETE instead of MUST_ERROR :
> Exception occurred while sending response. ExceptionMessage:Calling [asyncComplete()] is not valid for a request with Async state [MUST_ERROR]

That suggests issues with error handling. Either errors not being
handled or some sort of threading issue.

If you attempt to write to a closed stream Tomcat should simply swallow
the data and not send it to the client.

The only reason Tomcat will send a GOAWAY frame in normal usage if there
is a connection level error.

Mark

> 
> Thanks and Regards
> Arshiya Shariff
> 
> -----Original Message-----
> From: Martin Grigorov <mg...@apache.org> 
> Sent: Friday, September 25, 2020 4:37 PM
> To: Tomcat Users List <us...@tomcat.apache.org>
> Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR
> 
> On Fri, Sep 25, 2020 at 1:18 PM Arshiya Shariff <ar...@ericsson.com.invalid> wrote:
> 
>> Thanks Martin .
>>
>> Our expectation Is that , on receiving a RST_STREAM with CANCEL or 
>> NO_ERROR from the client after 1.5 secs for a particular stream , we 
>> don’t want the connection to be closed  with a GOAWAY:NO_ERROR  (while 
>> trying to send response after 2secs asynchronously over a stream that 
>> is closed) as the other streams during the same time are processing data fine .
>>
>> Additional info for GOAWAY: Connection [], Stream[],An error occurred 
>> during processing that was fatal to the connection .
>>
> 
> I guess some other error happened and that led to the GOAWAY. But I cannot be sure.
> Any errors in the logs ?
> 
> 
>>
>> Thanks and Regards
>> Arshiya Shariff
>>
>> -----Original Message-----
>> From: Martin Grigorov <mg...@apache.org>
>> Sent: Tuesday, September 22, 2020 7:31 PM
>> To: Tomcat Users List <us...@tomcat.apache.org>
>> Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a 
>> stream where the client has already sent RST_STREAM:NO ERROR
>>
>> Hi,
>>
>> On Tue, Sep 22, 2020 at 1:47 PM Arshiya Shariff < 
>> arshiya.shariff@ericsson.com.invalid> wrote:
>>
>>> Thank you so much Martin for the response.
>>>  Yes , 9.0.38 testing is on going .
>>>
>>> As we don’t get this clear with the RFC , please help us with the 
>>> below two cases :
>>
>>   * If a client sends RST_STREAM with NO_ERROR, then while sending 
>> async
>>> response is it expected behavior to close connection with GOAWAY ?
>>>   * If a client sends RST_STREAM with CANCEL , then while sending 
>>> async response will tomcat send RST_STREAM or GOAWAY , from http2 
>>> protocol perspective ?
>>>
>>
>> As per https://tools.ietf.org/html/rfc7540#section-6.4 and
>> https://tools.ietf.org/html/rfc7540#section-5.1 if "send R" or "recv R"
>> are processed then the stream is moved to CLOSED state.
>> GOAWAY should be sent to the client only if the connection will be 
>> closed ( https://tools.ietf.org/html/rfc7540#section-6.8). The server 
>> should close the connection only if some serious problem has happened, 
>> e.g. an IOException.
>>
>> Tell us more about your use case. What do you want to do when 1.5secs 
>> pass ? What do you expect to happen ?
>>
>>
>>>
>>> Thanks and Regards
>>> Arshiya Sharif
>>>
>>> -----Original Message-----
>>> From: Martin Grigorov <mg...@apache.org>
>>> Sent: Tuesday, September 22, 2020 1:18 AM
>>> To: Tomcat Users List <us...@tomcat.apache.org>
>>> Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over 
>>> a stream where the client has already sent RST_STREAM:NO ERROR
>>>
>>> Hi,
>>>
>>> On Mon, Sep 21, 2020 at 7:56 PM Arshiya Shariff < 
>>> arshiya.shariff@ericsson.com.invalid> wrote:
>>>
>>>> Hi All,
>>>>
>>>> The client has configured a response timeout of 1.5 seconds. In a 
>>>> case when our application tries to respond over a http2 stream 
>>>> asynchronously after 2 seconds where the client has already sent 
>>>> RST_STREAM with NO ERROR in 1.5 seconds
>>>
>>>
>>> Why does the client send NO_ERROR to the server ? I think it should 
>>> send a CANCEL instead.
>>> https://tools.ietf.org/html/rfc7540 mentions NO_ERROR only for 
>>> "Graceful shutdown of the server".
>>>
>>>
>>>> (due to no response) , then tomcat sends GOAWAY and closes the 
>>>> HTTP2 connection . Is this behavior of GOAWAY and connection 
>>>> closure
>> expected ?
>>>> We have planned to upgrade to Embedded tomcat version 9.0.38 . 
>>>> These are the behaviors we see in production with version 9.0.22 ,  
>>>> so we need some help with analyzing / validating  the existing 
>>>> behavior before
>>> the upgrade .
>>>> Please let us know.
>>>>
>>>
>>> Friendly advice:
>>> Please setup 9.0.38 locally and test on it.
>>> 9.0.22 is way too old. It is up to you to use it for your production 
>>> but for reporting bugs it is recommended to use the latest available
>> version.
>>> I, personally, prefer to spend my spare time with my family and 
>>> friends than to debug old versions just because the user doesn't 
>>> bother to test on a newer version.
>>>
>>>
>>>>
>>>> Thanks and Regards
>>>> Arshiya Shariff
>>>>
>>>
>>> --------------------------------------------------------------------
>>> - To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
>>> For additional commands, e-mail: users-help@tomcat.apache.org
>>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
>> For additional commands, e-mail: users-help@tomcat.apache.org
>>
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org


RE: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Posted by Arshiya Shariff <ar...@ericsson.com.INVALID>.
Thanks for the response Martin.

Below is the only exception we are getting in the logs while sending response to a closed stream , few times with MUST_COMPLETE instead of MUST_ERROR :
Exception occurred while sending response. ExceptionMessage:Calling [asyncComplete()] is not valid for a request with Async state [MUST_ERROR]

Thanks and Regards
Arshiya Shariff

-----Original Message-----
From: Martin Grigorov <mg...@apache.org> 
Sent: Friday, September 25, 2020 4:37 PM
To: Tomcat Users List <us...@tomcat.apache.org>
Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

On Fri, Sep 25, 2020 at 1:18 PM Arshiya Shariff <ar...@ericsson.com.invalid> wrote:

> Thanks Martin .
>
> Our expectation Is that , on receiving a RST_STREAM with CANCEL or 
> NO_ERROR from the client after 1.5 secs for a particular stream , we 
> don’t want the connection to be closed  with a GOAWAY:NO_ERROR  (while 
> trying to send response after 2secs asynchronously over a stream that 
> is closed) as the other streams during the same time are processing data fine .
>
> Additional info for GOAWAY: Connection [], Stream[],An error occurred 
> during processing that was fatal to the connection .
>

I guess some other error happened and that led to the GOAWAY. But I cannot be sure.
Any errors in the logs ?


>
> Thanks and Regards
> Arshiya Shariff
>
> -----Original Message-----
> From: Martin Grigorov <mg...@apache.org>
> Sent: Tuesday, September 22, 2020 7:31 PM
> To: Tomcat Users List <us...@tomcat.apache.org>
> Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a 
> stream where the client has already sent RST_STREAM:NO ERROR
>
> Hi,
>
> On Tue, Sep 22, 2020 at 1:47 PM Arshiya Shariff < 
> arshiya.shariff@ericsson.com.invalid> wrote:
>
> > Thank you so much Martin for the response.
> >  Yes , 9.0.38 testing is on going .
> >
> > As we don’t get this clear with the RFC , please help us with the 
> > below two cases :
>
>   * If a client sends RST_STREAM with NO_ERROR, then while sending 
> async
> > response is it expected behavior to close connection with GOAWAY ?
> >   * If a client sends RST_STREAM with CANCEL , then while sending 
> > async response will tomcat send RST_STREAM or GOAWAY , from http2 
> > protocol perspective ?
> >
>
> As per https://tools.ietf.org/html/rfc7540#section-6.4 and
> https://tools.ietf.org/html/rfc7540#section-5.1 if "send R" or "recv R"
> are processed then the stream is moved to CLOSED state.
> GOAWAY should be sent to the client only if the connection will be 
> closed ( https://tools.ietf.org/html/rfc7540#section-6.8). The server 
> should close the connection only if some serious problem has happened, 
> e.g. an IOException.
>
> Tell us more about your use case. What do you want to do when 1.5secs 
> pass ? What do you expect to happen ?
>
>
> >
> > Thanks and Regards
> > Arshiya Sharif
> >
> > -----Original Message-----
> > From: Martin Grigorov <mg...@apache.org>
> > Sent: Tuesday, September 22, 2020 1:18 AM
> > To: Tomcat Users List <us...@tomcat.apache.org>
> > Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over 
> > a stream where the client has already sent RST_STREAM:NO ERROR
> >
> > Hi,
> >
> > On Mon, Sep 21, 2020 at 7:56 PM Arshiya Shariff < 
> > arshiya.shariff@ericsson.com.invalid> wrote:
> >
> > > Hi All,
> > >
> > > The client has configured a response timeout of 1.5 seconds. In a 
> > > case when our application tries to respond over a http2 stream 
> > > asynchronously after 2 seconds where the client has already sent 
> > > RST_STREAM with NO ERROR in 1.5 seconds
> >
> >
> > Why does the client send NO_ERROR to the server ? I think it should 
> > send a CANCEL instead.
> > https://tools.ietf.org/html/rfc7540 mentions NO_ERROR only for 
> > "Graceful shutdown of the server".
> >
> >
> > > (due to no response) , then tomcat sends GOAWAY and closes the 
> > > HTTP2 connection . Is this behavior of GOAWAY and connection 
> > > closure
> expected ?
> > > We have planned to upgrade to Embedded tomcat version 9.0.38 . 
> > > These are the behaviors we see in production with version 9.0.22 ,  
> > > so we need some help with analyzing / validating  the existing 
> > > behavior before
> > the upgrade .
> > > Please let us know.
> > >
> >
> > Friendly advice:
> > Please setup 9.0.38 locally and test on it.
> > 9.0.22 is way too old. It is up to you to use it for your production 
> > but for reporting bugs it is recommended to use the latest available
> version.
> > I, personally, prefer to spend my spare time with my family and 
> > friends than to debug old versions just because the user doesn't 
> > bother to test on a newer version.
> >
> >
> > >
> > > Thanks and Regards
> > > Arshiya Shariff
> > >
> >
> > --------------------------------------------------------------------
> > - To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> > For additional commands, e-mail: users-help@tomcat.apache.org
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
>

Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Posted by Martin Grigorov <mg...@apache.org>.
On Fri, Sep 25, 2020 at 1:18 PM Arshiya Shariff
<ar...@ericsson.com.invalid> wrote:

> Thanks Martin .
>
> Our expectation Is that , on receiving a RST_STREAM with CANCEL or
> NO_ERROR from the client after 1.5 secs for a particular stream , we don’t
> want the connection to be closed  with a GOAWAY:NO_ERROR  (while trying to
> send response after 2secs asynchronously over a stream that is closed) as
> the other streams during the same time are processing data fine .
>
> Additional info for GOAWAY: Connection [], Stream[],An error occurred
> during processing that was fatal to the connection .
>

I guess some other error happened and that led to the GOAWAY. But I cannot
be sure.
Any errors in the logs ?


>
> Thanks and Regards
> Arshiya Shariff
>
> -----Original Message-----
> From: Martin Grigorov <mg...@apache.org>
> Sent: Tuesday, September 22, 2020 7:31 PM
> To: Tomcat Users List <us...@tomcat.apache.org>
> Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a
> stream where the client has already sent RST_STREAM:NO ERROR
>
> Hi,
>
> On Tue, Sep 22, 2020 at 1:47 PM Arshiya Shariff <
> arshiya.shariff@ericsson.com.invalid> wrote:
>
> > Thank you so much Martin for the response.
> >  Yes , 9.0.38 testing is on going .
> >
> > As we don’t get this clear with the RFC , please help us with the
> > below two cases :
>
>   * If a client sends RST_STREAM with NO_ERROR, then while sending async
> > response is it expected behavior to close connection with GOAWAY ?
> >   * If a client sends RST_STREAM with CANCEL , then while sending
> > async response will tomcat send RST_STREAM or GOAWAY , from http2
> > protocol perspective ?
> >
>
> As per https://tools.ietf.org/html/rfc7540#section-6.4 and
> https://tools.ietf.org/html/rfc7540#section-5.1 if "send R" or "recv R"
> are processed then the stream is moved to CLOSED state.
> GOAWAY should be sent to the client only if the connection will be closed
> ( https://tools.ietf.org/html/rfc7540#section-6.8). The server should
> close the connection only if some serious problem has happened, e.g. an
> IOException.
>
> Tell us more about your use case. What do you want to do when 1.5secs pass
> ? What do you expect to happen ?
>
>
> >
> > Thanks and Regards
> > Arshiya Sharif
> >
> > -----Original Message-----
> > From: Martin Grigorov <mg...@apache.org>
> > Sent: Tuesday, September 22, 2020 1:18 AM
> > To: Tomcat Users List <us...@tomcat.apache.org>
> > Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a
> > stream where the client has already sent RST_STREAM:NO ERROR
> >
> > Hi,
> >
> > On Mon, Sep 21, 2020 at 7:56 PM Arshiya Shariff <
> > arshiya.shariff@ericsson.com.invalid> wrote:
> >
> > > Hi All,
> > >
> > > The client has configured a response timeout of 1.5 seconds. In a
> > > case when our application tries to respond over a http2 stream
> > > asynchronously after 2 seconds where the client has already sent
> > > RST_STREAM with NO ERROR in 1.5 seconds
> >
> >
> > Why does the client send NO_ERROR to the server ? I think it should
> > send a CANCEL instead.
> > https://tools.ietf.org/html/rfc7540 mentions NO_ERROR only for
> > "Graceful shutdown of the server".
> >
> >
> > > (due to no response) , then tomcat sends GOAWAY and closes the HTTP2
> > > connection . Is this behavior of GOAWAY and connection closure
> expected ?
> > > We have planned to upgrade to Embedded tomcat version 9.0.38 . These
> > > are the behaviors we see in production with version 9.0.22 ,  so we
> > > need some help with analyzing / validating  the existing behavior
> > > before
> > the upgrade .
> > > Please let us know.
> > >
> >
> > Friendly advice:
> > Please setup 9.0.38 locally and test on it.
> > 9.0.22 is way too old. It is up to you to use it for your production
> > but for reporting bugs it is recommended to use the latest available
> version.
> > I, personally, prefer to spend my spare time with my family and
> > friends than to debug old versions just because the user doesn't
> > bother to test on a newer version.
> >
> >
> > >
> > > Thanks and Regards
> > > Arshiya Shariff
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> > For additional commands, e-mail: users-help@tomcat.apache.org
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
>

RE: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Posted by Arshiya Shariff <ar...@ericsson.com.INVALID>.
Thanks Martin .

Our expectation Is that , on receiving a RST_STREAM with CANCEL or NO_ERROR from the client after 1.5 secs for a particular stream , we don’t want the connection to be closed  with a GOAWAY:NO_ERROR  (while trying to send response after 2secs asynchronously over a stream that is closed) as the other streams during the same time are processing data fine .

Additional info for GOAWAY: Connection [], Stream[],An error occurred during processing that was fatal to the connection .

Thanks and Regards
Arshiya Shariff

-----Original Message-----
From: Martin Grigorov <mg...@apache.org> 
Sent: Tuesday, September 22, 2020 7:31 PM
To: Tomcat Users List <us...@tomcat.apache.org>
Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Hi,

On Tue, Sep 22, 2020 at 1:47 PM Arshiya Shariff <ar...@ericsson.com.invalid> wrote:

> Thank you so much Martin for the response.
>  Yes , 9.0.38 testing is on going .
>
> As we don’t get this clear with the RFC , please help us with the 
> below two cases :

  * If a client sends RST_STREAM with NO_ERROR, then while sending async
> response is it expected behavior to close connection with GOAWAY ?
>   * If a client sends RST_STREAM with CANCEL , then while sending 
> async response will tomcat send RST_STREAM or GOAWAY , from http2 
> protocol perspective ?
>

As per https://tools.ietf.org/html/rfc7540#section-6.4 and
https://tools.ietf.org/html/rfc7540#section-5.1 if "send R" or "recv R" are processed then the stream is moved to CLOSED state.
GOAWAY should be sent to the client only if the connection will be closed ( https://tools.ietf.org/html/rfc7540#section-6.8). The server should close the connection only if some serious problem has happened, e.g. an IOException.

Tell us more about your use case. What do you want to do when 1.5secs pass ? What do you expect to happen ?


>
> Thanks and Regards
> Arshiya Sharif
>
> -----Original Message-----
> From: Martin Grigorov <mg...@apache.org>
> Sent: Tuesday, September 22, 2020 1:18 AM
> To: Tomcat Users List <us...@tomcat.apache.org>
> Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a 
> stream where the client has already sent RST_STREAM:NO ERROR
>
> Hi,
>
> On Mon, Sep 21, 2020 at 7:56 PM Arshiya Shariff < 
> arshiya.shariff@ericsson.com.invalid> wrote:
>
> > Hi All,
> >
> > The client has configured a response timeout of 1.5 seconds. In a 
> > case when our application tries to respond over a http2 stream 
> > asynchronously after 2 seconds where the client has already sent 
> > RST_STREAM with NO ERROR in 1.5 seconds
>
>
> Why does the client send NO_ERROR to the server ? I think it should 
> send a CANCEL instead.
> https://tools.ietf.org/html/rfc7540 mentions NO_ERROR only for 
> "Graceful shutdown of the server".
>
>
> > (due to no response) , then tomcat sends GOAWAY and closes the HTTP2 
> > connection . Is this behavior of GOAWAY and connection closure expected ?
> > We have planned to upgrade to Embedded tomcat version 9.0.38 . These 
> > are the behaviors we see in production with version 9.0.22 ,  so we 
> > need some help with analyzing / validating  the existing behavior 
> > before
> the upgrade .
> > Please let us know.
> >
>
> Friendly advice:
> Please setup 9.0.38 locally and test on it.
> 9.0.22 is way too old. It is up to you to use it for your production 
> but for reporting bugs it is recommended to use the latest available version.
> I, personally, prefer to spend my spare time with my family and 
> friends than to debug old versions just because the user doesn't 
> bother to test on a newer version.
>
>
> >
> > Thanks and Regards
> > Arshiya Shariff
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org


Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Posted by Martin Grigorov <mg...@apache.org>.
Hi,

On Tue, Sep 22, 2020 at 1:47 PM Arshiya Shariff
<ar...@ericsson.com.invalid> wrote:

> Thank you so much Martin for the response.
>  Yes , 9.0.38 testing is on going .
>
> As we don’t get this clear with the RFC , please help us with the below
> two cases :

  * If a client sends RST_STREAM with NO_ERROR, then while sending async
> response is it expected behavior to close connection with GOAWAY ?
>   * If a client sends RST_STREAM with CANCEL , then while sending async
> response will tomcat send RST_STREAM or GOAWAY , from http2 protocol
> perspective ?
>

As per https://tools.ietf.org/html/rfc7540#section-6.4 and
https://tools.ietf.org/html/rfc7540#section-5.1 if "send R" or "recv R" are
processed then the stream is moved to CLOSED state.
GOAWAY should be sent to the client only if the connection will be closed (
https://tools.ietf.org/html/rfc7540#section-6.8). The server should close
the connection only if some serious problem has happened, e.g. an
IOException.

Tell us more about your use case. What do you want to do when 1.5secs pass
? What do you expect to happen ?


>
> Thanks and Regards
> Arshiya Sharif
>
> -----Original Message-----
> From: Martin Grigorov <mg...@apache.org>
> Sent: Tuesday, September 22, 2020 1:18 AM
> To: Tomcat Users List <us...@tomcat.apache.org>
> Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a
> stream where the client has already sent RST_STREAM:NO ERROR
>
> Hi,
>
> On Mon, Sep 21, 2020 at 7:56 PM Arshiya Shariff <
> arshiya.shariff@ericsson.com.invalid> wrote:
>
> > Hi All,
> >
> > The client has configured a response timeout of 1.5 seconds. In a case
> > when our application tries to respond over a http2 stream
> > asynchronously after 2 seconds where the client has already sent
> > RST_STREAM with NO ERROR in 1.5 seconds
>
>
> Why does the client send NO_ERROR to the server ? I think it should send a
> CANCEL instead.
> https://tools.ietf.org/html/rfc7540 mentions NO_ERROR only for "Graceful
> shutdown of the server".
>
>
> > (due to no response) , then tomcat sends GOAWAY and closes the HTTP2
> > connection . Is this behavior of GOAWAY and connection closure expected ?
> > We have planned to upgrade to Embedded tomcat version 9.0.38 . These
> > are the behaviors we see in production with version 9.0.22 ,  so we
> > need some help with analyzing / validating  the existing behavior before
> the upgrade .
> > Please let us know.
> >
>
> Friendly advice:
> Please setup 9.0.38 locally and test on it.
> 9.0.22 is way too old. It is up to you to use it for your production but
> for reporting bugs it is recommended to use the latest available version.
> I, personally, prefer to spend my spare time with my family and friends
> than to debug old versions just because the user doesn't bother to test on
> a newer version.
>
>
> >
> > Thanks and Regards
> > Arshiya Shariff
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
> For additional commands, e-mail: users-help@tomcat.apache.org
>

RE: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Posted by Arshiya Shariff <ar...@ericsson.com.INVALID>.
Thank you so much Martin for the response.
 Yes , 9.0.38 testing is on going .

As we don’t get this clear with the RFC , please help us with the below two cases :
  * If a client sends RST_STREAM with NO_ERROR, then while sending async response is it expected behavior to close connection with GOAWAY ?
  * If a client sends RST_STREAM with CANCEL , then while sending async response will tomcat send RST_STREAM or GOAWAY , from http2 protocol perspective ?

Thanks and Regards
Arshiya Sharif

-----Original Message-----
From: Martin Grigorov <mg...@apache.org> 
Sent: Tuesday, September 22, 2020 1:18 AM
To: Tomcat Users List <us...@tomcat.apache.org>
Subject: Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Hi,

On Mon, Sep 21, 2020 at 7:56 PM Arshiya Shariff <ar...@ericsson.com.invalid> wrote:

> Hi All,
>
> The client has configured a response timeout of 1.5 seconds. In a case 
> when our application tries to respond over a http2 stream 
> asynchronously after 2 seconds where the client has already sent 
> RST_STREAM with NO ERROR in 1.5 seconds


Why does the client send NO_ERROR to the server ? I think it should send a CANCEL instead.
https://tools.ietf.org/html/rfc7540 mentions NO_ERROR only for "Graceful shutdown of the server".


> (due to no response) , then tomcat sends GOAWAY and closes the HTTP2 
> connection . Is this behavior of GOAWAY and connection closure expected ?
> We have planned to upgrade to Embedded tomcat version 9.0.38 . These 
> are the behaviors we see in production with version 9.0.22 ,  so we 
> need some help with analyzing / validating  the existing behavior before the upgrade .
> Please let us know.
>

Friendly advice:
Please setup 9.0.38 locally and test on it.
9.0.22 is way too old. It is up to you to use it for your production but for reporting bugs it is recommended to use the latest available version.
I, personally, prefer to spend my spare time with my family and friends than to debug old versions just because the user doesn't bother to test on a newer version.


>
> Thanks and Regards
> Arshiya Shariff
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org


Re: HTTP2: Tomcat sends GOAWAY when trying to respond over a stream where the client has already sent RST_STREAM:NO ERROR

Posted by Martin Grigorov <mg...@apache.org>.
Hi,

On Mon, Sep 21, 2020 at 7:56 PM Arshiya Shariff
<ar...@ericsson.com.invalid> wrote:

> Hi All,
>
> The client has configured a response timeout of 1.5 seconds. In a case
> when our application tries to respond over a http2 stream asynchronously
> after 2 seconds where the client has already sent RST_STREAM with NO ERROR
> in 1.5 seconds


Why does the client send NO_ERROR to the server ? I think it should send a
CANCEL instead.
https://tools.ietf.org/html/rfc7540 mentions NO_ERROR only for "Graceful
shutdown of the server".


> (due to no response) , then tomcat sends GOAWAY and closes the HTTP2
> connection . Is this behavior of GOAWAY and connection closure expected ?
> We have planned to upgrade to Embedded tomcat version 9.0.38 . These are
> the behaviors we see in production with version 9.0.22 ,  so we need some
> help with analyzing / validating  the existing behavior before the upgrade .
> Please let us know.
>

Friendly advice:
Please setup 9.0.38 locally and test on it.
9.0.22 is way too old. It is up to you to use it for your production but
for reporting bugs it is recommended to use the latest available version.
I, personally, prefer to spend my spare time with my family and friends
than to debug old versions just because the user doesn't bother to test on
a newer version.


>
> Thanks and Regards
> Arshiya Shariff
>