Xend FD leak upon migrate receive

Xend FD leak upon migrate receive

[Steve Prochniak]

[-- Attachment #1.1: Type: text/plain, Size: 228 bytes --]

Xend leaks one fd each time a guest migrates to it's server.  I'm wondering if this is a known problem with a known fix, or if someone who knows that code better than me could point me in the right direction.  

 

Thanks.

[-- Attachment #1.2: Type: text/html, Size: 1592 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Xend FD leak upon migrate receive

[Ian Campbell]

On Wed, 2013-05-01 at 17:50 +0100, Steve Prochniak wrote:
> Xend leaks one fd each time a guest migrates to it’s server.  I’m
> wondering if this is a known problem with a known fix, or if someone
> who knows that code better than me could point me in the right
> direction.  

I'm afraid xend is basically unmaintained these days (which has caused
to become deprecated in current releases) AFAIK there aren't any people
around who know the code in any great detail.

I don't recall having heard of anything along the lines of what you
report so I don't think it is a known problem.

Ian.



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Xend FD leak upon migrate receive

[Sander Eikelenboom]

Thursday, May 2, 2013, 1:42:11 PM, you wrote:

> On Wed, 2013-05-01 at 17:50 +0100, Steve Prochniak wrote:
>> Xend leaks one fd each time a guest migrates to it’s server.  I’m
>> wondering if this is a known problem with a known fix, or if someone
>> who knows that code better than me could point me in the right
>> direction.  

> I'm afraid xend is basically unmaintained these days (which has caused
> to become deprecated in current releases) AFAIK there aren't any people
> around who know the code in any great detail.

> I don't recall having heard of anything along the lines of what you
> report so I don't think it is a known problem.

Isn't this the same as the:
 test-amd64-i386-xend-winxpsp3 16 leak-check/check             fail  never pass
 test-amd64-i386-xend-qemut-winxpsp3 16 leak-check/check        fail never pass

 Reported in about every automated test (a bit hidden by ever present "guest-stop" problem
 (due to lack of the -F on shutdown for HVM guests without PV drivers))

 --
 Sander

> Ian.






_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Xend FD leak upon migrate receive

[Ian Campbell]

On Thu, 2013-05-02 at 14:36 +0100, Sander Eikelenboom wrote:
> Thursday, May 2, 2013, 1:42:11 PM, you wrote:
> 
> > On Wed, 2013-05-01 at 17:50 +0100, Steve Prochniak wrote:
> >> Xend leaks one fd each time a guest migrates to it’s server.  I’m
> >> wondering if this is a known problem with a known fix, or if someone
> >> who knows that code better than me could point me in the right
> >> direction.  
> 
> > I'm afraid xend is basically unmaintained these days (which has caused
> > to become deprecated in current releases) AFAIK there aren't any people
> > around who know the code in any great detail.
> 
> > I don't recall having heard of anything along the lines of what you
> > report so I don't think it is a known problem.
> 
> Isn't this the same as the:
>  test-amd64-i386-xend-winxpsp3 16 leak-check/check             fail  never pass
>  test-amd64-i386-xend-qemut-winxpsp3 16 leak-check/check        fail never pass

Why do you think that?

AFAIK this is checking for leaked files (not file descriptors) and
xenstore entries etc. Also it seems to pass for other test sequences
which also involve xend and migration.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Xend FD leak upon migrate receive

[Steve Prochniak]

Fyi - the leak occurs on the receiver side and only exists when using SSL migration.  I don't know if that rings any bells or not.

-----Original Message-----
From: Ian Campbell [mailto:Ian.Campbell@citrix.com] 
Sent: Thursday, May 02, 2013 9:42 AM
To: Sander Eikelenboom
Cc: Steve Prochniak; xen-devel@lists.xen.org
Subject: Re: [Xen-devel] Xend FD leak upon migrate receive

On Thu, 2013-05-02 at 14:36 +0100, Sander Eikelenboom wrote:
> Thursday, May 2, 2013, 1:42:11 PM, you wrote:
> 
> > On Wed, 2013-05-01 at 17:50 +0100, Steve Prochniak wrote:
> >> Xend leaks one fd each time a guest migrates to it's server.  I'm
> >> wondering if this is a known problem with a known fix, or if someone
> >> who knows that code better than me could point me in the right
> >> direction.  
> 
> > I'm afraid xend is basically unmaintained these days (which has caused
> > to become deprecated in current releases) AFAIK there aren't any people
> > around who know the code in any great detail.
> 
> > I don't recall having heard of anything along the lines of what you
> > report so I don't think it is a known problem.
> 
> Isn't this the same as the:
>  test-amd64-i386-xend-winxpsp3 16 leak-check/check             fail  never pass
>  test-amd64-i386-xend-qemut-winxpsp3 16 leak-check/check        fail never pass

Why do you think that?

AFAIK this is checking for leaked files (not file descriptors) and
xenstore entries etc. Also it seems to pass for other test sequences
which also involve xend and migration.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: Xend FD leak upon migrate receive

[Jim Fehlig]

Steve Prochniak wrote:
> Fyi - the leak occurs on the receiver side and only exists when using SSL migration.  I don't know if that rings any bells or not.
>   

Maybe op_quit() in tools/python/xen/xend/server/relocate.py needs to
handle cleaning up the pipe created in op_sslreceive()?

Regards,
Jim

> -----Original Message-----
> From: Ian Campbell [mailto:Ian.Campbell@citrix.com] 
> Sent: Thursday, May 02, 2013 9:42 AM
> To: Sander Eikelenboom
> Cc: Steve Prochniak; xen-devel@lists.xen.org
> Subject: Re: [Xen-devel] Xend FD leak upon migrate receive
>
> On Thu, 2013-05-02 at 14:36 +0100, Sander Eikelenboom wrote:
>   
>> Thursday, May 2, 2013, 1:42:11 PM, you wrote:
>>
>>     
>>> On Wed, 2013-05-01 at 17:50 +0100, Steve Prochniak wrote:
>>>       
>>>> Xend leaks one fd each time a guest migrates to it's server.  I'm
>>>> wondering if this is a known problem with a known fix, or if someone
>>>> who knows that code better than me could point me in the right
>>>> direction.  
>>>>         
>>> I'm afraid xend is basically unmaintained these days (which has caused
>>> to become deprecated in current releases) AFAIK there aren't any people
>>> around who know the code in any great detail.
>>>       
>>> I don't recall having heard of anything along the lines of what you
>>> report so I don't think it is a known problem.
>>>       
>> Isn't this the same as the:
>>  test-amd64-i386-xend-winxpsp3 16 leak-check/check             fail  never pass
>>  test-amd64-i386-xend-qemut-winxpsp3 16 leak-check/check        fail never pass
>>     
>
> Why do you think that?
>
> AFAIK this is checking for leaked files (not file descriptors) and
> xenstore entries etc. Also it seems to pass for other test sequences
> which also involve xend and migration.
>
> Ian.
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
>   

Re: Xend FD leak upon migrate receive

[Steve Prochniak]

Thanks Jim - 

That's really close to what I've been testing.

I close the p2cread fd in op_sslreceive (but outside of the except: scope), and I believe that the p2cwrite fd gets closed as part of the recv2fd thread that runs in connection.py.  So far so good.

Steve

-----Original Message-----
From: Jim Fehlig [mailto:jfehlig@suse.com] 
Sent: Tuesday, May 07, 2013 11:31 AM
To: Steve Prochniak
Cc: Sander Eikelenboom; Ian Campbell; xen-devel@lists.xen.org
Subject: Re: [Xen-devel] Xend FD leak upon migrate receive

Steve Prochniak wrote:
> Fyi - the leak occurs on the receiver side and only exists when using SSL migration.  I don't know if that rings any bells or not.
>   

Maybe op_quit() in tools/python/xen/xend/server/relocate.py needs to
handle cleaning up the pipe created in op_sslreceive()?

Regards,
Jim

> -----Original Message-----
> From: Ian Campbell [mailto:Ian.Campbell@citrix.com] 
> Sent: Thursday, May 02, 2013 9:42 AM
> To: Sander Eikelenboom
> Cc: Steve Prochniak; xen-devel@lists.xen.org
> Subject: Re: [Xen-devel] Xend FD leak upon migrate receive
>
> On Thu, 2013-05-02 at 14:36 +0100, Sander Eikelenboom wrote:
>   
>> Thursday, May 2, 2013, 1:42:11 PM, you wrote:
>>
>>     
>>> On Wed, 2013-05-01 at 17:50 +0100, Steve Prochniak wrote:
>>>       
>>>> Xend leaks one fd each time a guest migrates to it's server.  I'm
>>>> wondering if this is a known problem with a known fix, or if someone
>>>> who knows that code better than me could point me in the right
>>>> direction.  
>>>>         
>>> I'm afraid xend is basically unmaintained these days (which has caused
>>> to become deprecated in current releases) AFAIK there aren't any people
>>> around who know the code in any great detail.
>>>       
>>> I don't recall having heard of anything along the lines of what you
>>> report so I don't think it is a known problem.
>>>       
>> Isn't this the same as the:
>>  test-amd64-i386-xend-winxpsp3 16 leak-check/check             fail  never pass
>>  test-amd64-i386-xend-qemut-winxpsp3 16 leak-check/check        fail never pass
>>     
>
> Why do you think that?
>
> AFAIK this is checking for leaked files (not file descriptors) and
> xenstore entries etc. Also it seems to pass for other test sequences
> which also involve xend and migration.
>
> Ian.
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
>   

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel