From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.3a) Gecko/20021207 Phoenix/0.5 Description of problem: Example machines were stock-installed with RHAS 2.1, upgraded to latest RHAS .rpms (as of 1/20/2003) via RHN, running 2.4.9-e8 series kernels. Hardware is compaq dl380-G2, 4 gB RAM, though we've reproduced it on different hardware, too. Both machines are on same switch, same VLAN, both machines locked to 100-full-duplex on switch and machines. No ethernet collisions on switch or machines. Both machines are using on-board Intel ethernet controllers: 00:02.0 Ethernet controller: Intel Corporation 82557 [Ethernet Pro 100] (rev 08) However, the problem appears using every NIC we've tried (3com 3c996SX gig-e, 3c985 gig-e). scp (using default RHAS config files in /etc/ssh, and no commandline options) starts at normal rate of 8-9 mB/s, then slows to a stop after about 500 mB of copying. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. dd if=/dev/zero of=/opt/2gB-file bs=1024 count=2M 2. scp 2gB-file <any other RHAS machine>:/opt/ Actual Results: scp slows & eventually makes little/no progress Expected Results: consistent copy at 7-10 mB/s. All our RH 7.1 & 7.3 machines do this.
Not only does the file copy slow down, but I've had files finish corrupt after taking more than four hours to copy. We are unable to copy ISO images from machine to machine because of this.
Could you try sftp? Or recompiling a newer openssh .src.rpm from Fedora?
Created attachment 114774 [details] Don't disable the nagle algorithm Could you compile openssh with this patch applied if it makes any difference?
No response.