high latencies on routing interface on Powerconnect 6248

Hi,

I checked the arp table and there are no Multicast MAC addresses (any MAC addresses beginning with 1)

i set the arp timeout to 300 secs.  I also took your suggection and did a no arp dynamicrenew.  This took my arp entries down to 358 from 545 entries. 

We are not using MS Load Balancing Clusters anywhere on the network.

We also dont have any multipaths from the uplinked switches to the Dell PowerConnect ...all connections to this PowerConnect are uplinks (no hosts) from other switches via one uplink cable.

i also checked and there is no multipaths from servers to switch.  The switch that connects the servers is uplinked to the Dell PowerConnect.

I am still seeing high latencies and and random packet loss......

Any other suggections to look for....I am beginning to get really frustrated with this device...

Thanks Much

Darn.

Well, let's go back to my default root cause of most problems: "flow control" or "spanning tree"

Flow control:

PowerConnect switches ship with flow control disabled.  Most other switches and all client and servers ship with flow control enabled.  Enabling this on all switches and end stations in your network will prevent the switch from dropping packets during times of over subscription.  Please enable flow control on all the switches in your network and double check the end stations that they have flow control enabled also.  I do not think this is your problem, but let's get it out of the way if we can.

console(config)# flowcontrol

Spanning tree:

Something in spanning tree is more likely to be your problem.  Your log file shows many topology changes.  You should not see any topology changes in a stable network, so this is a clue.   Are all the edge ports (ports not connected to switches) set to portfast?  What other switches (brand and model) are in your network?  Are any of them running Cisco's PVST?  What other switches (brand and model) are directly connected to the 62xx stack?  Please post a new (from today) "show logging" output?

Another thing to try, this is just a hunch, turn off remote logging and see if this improves the routing performance. 

console(config)# no logging x.x.x.x

I still think STP is the main issue, so make every effort to answer the questions in the previous post.

I temporarily shut off the remote logging and will keep an eye out for these problems.....will let you know after I let it run for a day.....

Thanks

Hi, The other switches connected to this switch are Dell Powerconnect 5448's except 2 ehich are cisco 2960G's.

I checked STP on the cisco's and they only support PVST.........

Flow Control I never turned on...but i can try it...i dont think i ever used Flow Control and not sure exactly what i should expect using Flow Control.  Can you enlighten me what flow control is used for?

Thanks

When an Ethernet device gets over loaded, flow control allows it to send PAUSE requests to the directly connected devices that are sending it data.  These devices then pause in sending data to allow the over loaded condition time to clear.  If flow control is not enabled and an over loaded condition occurs, the device will drop packets.   Dropping packets is much more performance impacting than flow control.

Again, I believe the root of your problem is spanning tree.  Specifically all the topology change notifations indicated in your log file.  These can be caused by ports in the network coming up.   For instance, when a laptop users connect to the network or when workstations power up in the morning or a device constantly rebooting.  These ports need to be set to portfast to prevent this.

Regarding the Cisco switches running PVST.  This could also be a problem.  Standards based switches, like PowerConnect, will simply forward PVST traffic as multicast traffic.    Depending how they are connected to the network, this can confuse the Cisco switches.  How are the Cisco switches connected into the network in relation to the 62xx switches? 

Any update?  Did turning off remote logging or the STP changes help? 

Any update?

Hi there

I'm having similar issues, there's always this occasional packet that's getting dropped.

I've setup 2 6224s and 1 6248. Each servers a location that's connected via fiber. Each one is the default gateway for each site and I'm only allowing L3 to pass between.

There are quite a lot more VLANs on then you have in your setup.

STP used to be a factor but STP isn't passed between sites any more and topology hardly ever changes still I'm loosing a packet every now and then.

I noticed that the DELLs have "Proxy Arp" enbled by default on each VLAN interface, any one done any performance testing with or without it?

Thanks,

Simon

I am waiting on the go ahead to turn on flow control...I anticipate this to happen Wednesday morning.....

Tempar2K,

Please start a new thread on your question.

Hi there..

I'd suggest like a previous posted did to check your STP topology.

Ensure you're using MSTP or at least RSTP.

Put all user ports in Port Fast and Guard Root

Put your core switch in a higher priority in STP, (lower number is higher Pri)

I've never see STP shake switches as much as with these Dells that I've recenetly been working on.

Please post the results of your changes.  Are you also going to put all ports not connected to other switches in portfast?

Were able to make the config file changes?  Did they help?  What are you seeing now?