FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > CentOS > CentOS

 
 
LinkBack Thread Tools
 
Old 12-22-2009, 12:06 PM
Sean Carolan
 
Default SFTP - stalled - on large files

> Load balancer... is that set up to maintain connections, or will it, like IBM's
> WebSeal, go to whichever server is next/least used in the middle of a connection?

It's set to use "least connection" but there is only one server behind
the virtual IP at the moment.

I'm reasonably sure at this point that the Netscaler is causing the
problem, because file transfers inside the LAN work fine, and we see
this same issue on both physical and virtual servers. I just tested
with a physical box to verify, and the same thing happens, transfer
speed quickly drops to zero and stalls.

I've got a ticket open with Citrix to hopefully get to the bottom of
this. It wouldn't be the first time we've seen the Netscaler muck up
a TCP connection from a client. The last time I dealt with this it
was sending unwanted FIN packets to mail servers. Fun stuff.
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 12-22-2009, 11:41 PM
Jay Leafey
 
Default SFTP - stalled - on large files

We had a similar problem copying files between servers on two of our
campuses via SCP. After a while the connection just stalled out and
hung. The problem turned out to be SCP and SFTP interacting a bug in
the SACK (Selective Acknowledgment) algorithm used in Linux. We turned
it off on the two endpoints using the following addition to
/etc/sysctl.conf:



# Turn off SACK
net.ipv4.tcp_sack = 0


and execute "sysctl -p" to apply it. You can also use "sysctl -w
net.ipv4.tcp_sack=0" to turn it off temporarily. Our file transfers
worked just fine after the change.


I realize there are differences our situation and yours and this might
not work in your case. Given the length of this thread, though, it
might be worth a try!

--
Jay Leafey - Memphis, TN
jay.leafey@mindless.com
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 12-27-2009, 02:32 PM
Sean Carolan
 
Default SFTP - stalled - on large files

>> # Turn off SACK
>> net.ipv4.tcp_sack = 0
>
> and execute "sysctl -p" to apply it. *You can also use "sysctl -w
> net.ipv4.tcp_sack=0" to turn it off temporarily. *Our file transfers worked
> just fine after the change.
>
> I realize there are differences our situation and yours and this might not
> work in your case. *Given the length of this thread, though, it might be
> worth a try!

It appears that the Netscaler load balancer was the problem. We
turned off TCP buffering (TCPB) on the netscaler for this particular
virtual server, and I was immediately able to transfer a 95MB file
with no issues. Citrix has acknowledged that there may be some issues
with the tcp stack on this device, which they think have been resolved
in more recent versions of the Netscaler OS.

Hopefully if anyone else experiences this issue, they'll be able to
Google it via the CentOS list archives.
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 

Thread Tools




All times are GMT. The time now is 03:18 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org