0

I have 2 RedHat servers (say rhel-01 and rhel-02) with 40G ports connected via 40G switch. When I'm trying to copy any large file (larger than 1G) from rhel-01 to rhel-02 it stalls at exactly 98% for one to one and half minutes and then completes the job.

[root@rhel-01 ~]# cd /data1/
[root@rhel-01 data1]# scp 1G user@192.168.100.12:/home/user
user@192.168.100.12's password:
1G              100% 1024MB 693.2MB/s   00:01
[root@rhel-01]# scp 10G user@192.168.100.12:/home/user
user@192.168.100.12's password:
10G             100%   10GB 198.6MB/s   00:51
[root@CGP-ICMS-Hadoop-Node1 data1]# scp 20G user@192.168.100.12:/home/user
user@192.168.100.12's password:
20G             100%   20GB 194.7MB/s   01:45
[root@CGP-ICMS-Hadoop-Node1 data1]#

I have no idea what could be the issue. Can anyone help me out here, please?

akjon RooSho
  • 1

1 Answers1

1

The server is probably busy writing to disk.

SCP doesn't finish until data is actually written to disk. You have a fast network, so the disks can probably not keep up with the flow of data, and thus speed drops when the remote ends buffer is full and it has to write data to disk.

vidarlo
  • 6,654
  • 2
  • 18
  • 31
  • Is there any way to diagnose the issue? – akjon RooSho Jul 03 '22 at 16:24
  • Look for thread state on the remote end. It'll probably be waiting. In addition, you should check disk performance. The fact that your speed decreases with increasing filesize indicates that disk, and not network, is your bottleneck. – vidarlo Jul 03 '22 at 16:42