Connection reset by peer

Connection reset by peer

[Joern Rennecke]

My regression test checkouts fail, with messages like:
cvs [checkout aborted]: recv() from server gcc.gnu.org: Connection reset by peer
and
cvs [checkout aborted]: recv() from server sources.redhat.com: Connection reset by peer
.

Re: Connection reset by peer

[Angela Marie Thomas]

> bash$ host gcc.gnu.org
> gcc.gnu.org has address 209.249.29.67
> gcc.gnu.org mail is handled (pri=0) by gcc.gnu.org

Since the load was next to nil, I did some bashing on both IP addresses
and can only get failures on the new one.  There's quite a lot of outgoing
traffic (ftp, http it looks like).  It's possible that the kernel just
can't handle such a big pipe and is dropping connections when it's
overloaded.  It looks like I get a failed connection about 1/3 of the
time.  I'll see if I can increase the logging and get some more info
for our networking guy when he wakes up (he was the one who moved it
so was up late).

--Angela

Re: Connection reset by peer

[Angela Marie Thomas]

> Actually, that didn't work either: "no space on device" in the
> gccadmin directory.  Is there a partition with lots of free space
> for me to play in for a few hours, or are we just out of luck?

/sourceware/snapshot-tmp has gobs of space and is I believe intended
for precisely this sort of thing.

--Angela

Re: Connection reset by peer

[Mark Mitchell]

> It would be most helpful to know which IP address you are accessing when
> this happens.  Can you please see which IP address your nameserver is
> resolving for gcc.gnu.org?

bash$ host gcc.gnu.org
gcc.gnu.org has address 209.249.29.67
gcc.gnu.org mail is handled (pri=0) by gcc.gnu.org

Thanks for following up.

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: Connection reset by peer

[Mark Mitchell]

--On Sunday, August 19, 2001 12:53:04 PM -0700 Jason Molenda 
<jason@molenda.com> wrote:

> On Sun, Aug 19, 2001 at 12:40:57PM -0700, Mark Mitchell wrote:
>
>> morning?  Unfortunately, I'm trying to do a GCC release, and this
>> is making it difficult.  I don't want to run the script on gcc.gnu.org
>> directly since that will consume a lot of CPU.
>>
>
> Unless a fix pops up right now, do it on the server.  Today is a
> Sunday and system load is going to be very light.

Actually, that didn't work either: "no space on device" in the
gccadmin directory.  Is there a partition with lots of free space
for me to play in for a few hours, or are we just out of luck?

Thanks,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: Connection reset by peer

[Mark Mitchell]

--On Sunday, August 19, 2001 12:53:04 PM -0700 Jason Molenda 
<jason@molenda.com> wrote:

> On Sun, Aug 19, 2001 at 12:40:57PM -0700, Mark Mitchell wrote:
>
>> morning?  Unfortunately, I'm trying to do a GCC release, and this
>> is making it difficult.  I don't want to run the script on gcc.gnu.org
>> directly since that will consume a lot of CPU.
>>
>
> Unless a fix pops up right now, do it on the server.  Today is a
> Sunday and system load is going to be very light.

Roger, wilco.

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: Connection reset by peer

[Angela Marie Thomas]

> Does anyone know why I am gettinga  lot of messages like this:
> 
> read: Connection reset by peer
> cvs [commit aborted]: end of file from server (consult above messages if 
> any)

Since this is the third complaint, I've put a call into our networking
guy to verify that everything is OK.  I can't find anything in the logs
to indicate anything obvious.  After about 20 tries, I finally got one
of my cvs commands to fail.

It would be most helpful to know which IP address you are accessing when
this happens.  Can you please see which IP address your nameserver is
resolving for gcc.gnu.org?

Thanks

--Angela

Re: Connection reset by peer

[Jason Molenda]

On Sun, Aug 19, 2001 at 12:40:57PM -0700, Mark Mitchell wrote:

> morning?  Unfortunately, I'm trying to do a GCC release, and this
> is making it difficult.  I don't want to run the script on gcc.gnu.org
> directly since that will consume a lot of CPU.
> 

Unless a fix pops up right now, do it on the server.  Today is a
Sunday and system load is going to be very light.

Jason

Connection reset by peer

[Mark Mitchell]

Does anyone know why I am gettinga  lot of messages like this:

read: Connection reset by peer
cvs [commit aborted]: end of file from server (consult above messages if 
any)

messages when trying to check things into the GCC CVS repository this
morning?  Unfortunately, I'm trying to do a GCC release, and this
is making it difficult.  I don't want to run the script on gcc.gnu.org
directly since that will consume a lot of CPU.

Thanks,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com