Description of problem: RHR2 net test fails with "Connection reset by peer" It looks to me like the SSH connection is what is failing, but I cannot determine why. Version-Release number of selected component (if applicable): rhr2-rhel4-1.0-10 How reproducible: Everytime Steps to Reproduce: 1. 2. 3. Actual results: Test fails Expected results: Test passes Additional info: Failure is reproducable both with a static and a DHCP ip address.
Created attachment 110155 [details] NETWORK/output.log
Created attachment 110156 [details] NETWORK/hardware.log
Raising severity to HIGH. I tried to find a severity of BLOCKER, because this is blocking our ability to certify the hardware, but that severity is not in the list. An issue tracker will be opened shortly.
What version of RHEL are you running on the iSpec server? This may be the same as bug 145570. Can you try the change described in that ticket (change the ab parameters) and see if that addresses the problem?
The iSpec server is running RHEL4 prec-RC2. We can upgrade it if need be, but we looked at the RPM versions of both Apache and SSH between the pre-rc2 and the final and couldn't find any difference. We will test the work-around by modifying the 'ab' command in /usr/share/rhr/tests/network/tcp as suggested in bug 145570.
I agree; I do not think that upgrading the iSpec server will change anything. Let us know how the ab change goes.
I am sorry to report that the ab change did not help us. We are still receiving the "Connection reset by peer" error. We've now run the test against an Integrated e100 NIC and that one passes (but for some reasons its name is dev6536 and not ethx). But the test still fails when run against our gigabit Intel controllers which use the e1000 driver. I will post a new log file set shortyl.
Created attachment 110252 [details] output.log from e100 (PASS) and e1000 (FAIL)
Created attachment 110253 [details] hardware.log from e100 (PASS) and e1000 (FAIL)
Can the ab line be run by hand against the e1000 NIC and pass or does that always fail?
Yes, the ab command was able to pass "by hand" against the e1000 NIC. This was by running it from the ispec server and not through an SSH command. I believe the SSH was timing out. A good reason for this is that while our SUT had a gigabit connection, the ispec server did not. So ab was downloading a 125MB file over a 100mbit connetion. It would take atleast 15 minutes for text "Processed 100 Requests" to be displayed to the screen. I think the lack of interaction (display and typing) caused the SSH connection to time out. We have since moved ispec to a newer server that supports gigabit. We rerun the test and it passes as expected. Thak you for all your help, and I apologize for our error when setting up our environment.