So what's connected to the 2716? Servers and client workstations only? I'm trying to get an idea of what the network looks like. Do you have clients & servers connected directly to the 2716 or other stuff in your network. Is the host that's trying to get to the server going through multiple different hops before it get to the 2716? Can you describe your network better?
What version of the FW are you running exactly?
Can you check the port configuration of any of the devices or switch ports or whatever that's connected to the 2716 and see if the ports are configured as "fix speed/duplex" or "auto-negotiation"? If one side of the link is fixed speed/duplex and the other side is auto then you may cause problems across the link which could result in packet drops. This could happen on any link so check all the links from the servers through the switches to the clients.
Do you experience lost only when you are doing backup? So does normal operation works ok for you? Can you try sending ping with varying packet sizes to see if you experience lost only on large packet sizes? There was a known problem with Jumbo frame earlier which could explain this problem too if you are seeing problems only on large packets - are you running jumbo frame on your network?
Right now only a couple of test servers are connected to the 2716 because of the problems we were having with it. This is in a university distributed network. We are a department with a /23 VLAN on that network. The first time i tested the switch, it was in the same room and on the same main switch as the backup server. Now it is in another building. Both with the same results.
Firmware: 1.0.1.07
Boot: 1008
Thats the easiest way to tell so far that it is losing its connection is that the backups are failing on the servers connected to the 2716 randomly. If i take out the 2716 and replace it with another brand switch, the backups works perfect.
Ports are all set to auto. No Jumboframes. Again, it did the same thing in the un-managed mode as well. Pings of various sizes and times go through fine.
That's very strange. There was a known issue before where the switch was incorrectly calculating packets size when VLAN is enabled and large files were being transferred, so that it mistakenly drop packets that were close to max MTU because it thought that the packet was actually a jumbo size. This was corrected however in the latest code. Now this problem would cause packets to get dropped but the link would not go down.
When you said that your connection is lost, is it permanently lost or do you mean packet drop? So when you notice that the connection is lost did you do something to regain connection or does it comes back on its own? Is it possible that it is packet lost you are experiencing or does the link actually goes down?
To eliminate other equipments in the mix, is it possible to connect two hosts (server & client) directly to the 2716 and just do a backup to see if it fails when running standalone (not connected to other switches)? I'm just trying to understand if there is some interaction between the 2716 and other network equipments?
Arcserve slows to a crawl on the host, then it'll disconnect with a connection to host lost. The backups take place in the wee hours of the morning, but when i check it when i come into work, the connection is fine.
I can try and do a windows backup to another host on the same switch, but it wouldn't really replicate the conditions.
This sure sounds like the same problem as we observed before but it should be fixed if you are running latest code. One way to see if it is related is to do this - just enable "jumbo frame" on this 2716 switch (don't enable jumbo frame anywhere else) - just enable it on the 2716 switch. What this does is to cause the 2716 to pass through the mis-calculated frames. If this works then it would seem that the problem is still related to the issue we found before.
You should probably test this standalone so it won't interfere with normal operation.
DELL-Cuong N.
1K Posts
0
April 17th, 2006 17:00
vtek63
23 Posts
0
April 17th, 2006 18:00
DELL-Cuong N.
1K Posts
0
April 17th, 2006 18:00
That's very strange. There was a known issue before where the switch was incorrectly calculating packets size when VLAN is enabled and large files were being transferred, so that it mistakenly drop packets that were close to max MTU because it thought that the packet was actually a jumbo size. This was corrected however in the latest code. Now this problem would cause packets to get dropped but the link would not go down.
When you said that your connection is lost, is it permanently lost or do you mean packet drop? So when you notice that the connection is lost did you do something to regain connection or does it comes back on its own? Is it possible that it is packet lost you are experiencing or does the link actually goes down?
To eliminate other equipments in the mix, is it possible to connect two hosts (server & client) directly to the 2716 and just do a backup to see if it fails when running standalone (not connected to other switches)? I'm just trying to understand if there is some interaction between the 2716 and other network equipments?
Cuong.
vtek63
23 Posts
0
April 17th, 2006 19:00
DELL-Cuong N.
1K Posts
0
April 17th, 2006 19:00
This sure sounds like the same problem as we observed before but it should be fixed if you are running latest code. One way to see if it is related is to do this - just enable "jumbo frame" on this 2716 switch (don't enable jumbo frame anywhere else) - just enable it on the 2716 switch. What this does is to cause the 2716 to pass through the mis-calculated frames. If this works then it would seem that the problem is still related to the issue we found before.
You should probably test this standalone so it won't interfere with normal operation.
Cuong.
vtek63
23 Posts
0
April 17th, 2006 20:00