Slow/unstable performance on 5300xl

[IT_7]

We've had this problem before and after RSTP operation.

[Thilo Schüler-Mach]

Hi all,
at this week i finished the physical connections of our new hp switches, 2x 530x, 3x 28xx. They are all connected by single ge-links or ge-trunks. So at this morning i want to do some quick tests befor i shutdown our cisco switches and switch all our devices to hp. Convergence by MSTP works fine in our littel ring topology, we have no problem.
But the Inter-VLAN-Routing test results were horrible.
We use both sfp-ports on on 530x module and send packet sizes from 64 to 1500 byte from 253 sources to 253 destinations.
Without ACL and QoS we get a max. performance of 5 MBit/s on bidirectional and max. 8 MBit/s on unidirectonal traffic by 50% - 60 % CPU load. If we increase the load, the switch starts to drop packests and the available bandwidth decrease to 1-2 MBit/s by 30% - 35% CPU load.
Switching performances lookes good by 280 MBit/s with 64 byte packets bidir. traffic and 960 MBit/s with 64 - 1500 byte packets bidir. traffic.
So witch mistake i have made in such a simple routing scenario?
FYI we use the current 9.22 image, mstp, gvrp and 256 vlans.

Thanks for any help,
Thilo
r

[Carl Bralick_1]

We finally had to disable spanning tree to clear up the performance problems. We spent 2 weeks troubleshooting with HP Tech Support and they insisted we had a loop. I personally traced each and every fiber strand on our campus to ensure that wasn't so. It is up and running fine now. Prior to turning off spanning tree, I upgraded the firmware to version 9.22 - this had no affect.

[Thilo Schüler-Mach]

Hi Carl,
to decide between routing or spanning-tree isn`t realy a choice, isn´t it?

regards
Thi

[Bjorn Tore Paulen]

We did use RSTP in our network. HP claims - at least when there is a problem - that the draft states there can be only 7 switches in a STP domain. We saw that even though nothing was done in the network, our switches recalculated the STP domain every now and then. I actually have one switch in my network that recalculates every second. This, I think, is due to customer's equipment connected to it. However, recalculations should perform every time something is connected/disconnected to the STP domain. Our customers lost their VoIP sessions during recalculation; would be syrup for ~4 sec.

This is a healthy switch, not connected to any other STP enabled equipment. Note the "Time since last change" and "Topology change count":

Rapid Spanning Tree (RSTP) Information

STP Enabled : Yes
Force Version : RSTP-operation

Switch Priority : 32768 Hello Time : 2
Max Age : 20 Forward Delay : 15

Topology Change Count : 9
Time Since Last Change : 35 hours

Root MAC Address : 000a57-04bc00
Root Path Cost : 0
Root Port : This switch is root
Root Priority : 32768


This is a not that healthy switch, running 9.22:

Rapid Spanning Tree (RSTP) Information

STP Enabled : Yes
Force Version : RSTP-operation

Switch Priority : 0 Hello Time : 2
Max Age : 20 Forward Delay : 15

Topology Change Count : 11,127
Time Since Last Change : 1 secs

Root MAC Address : 000e7f-bd3700
Root Path Cost : 0
Root Port : This switch is root
Root Priority : 0



...go figure..

[stieven struyf]

We are also having the same issues(on a brandnew 5312zl). Although sometimes it takes 2-3hours to start after a complete reboot of the switch. We were not even using the switch as a router.
We had to go back after first migration attempt. We had several engineer of different companies(network/firewall experts, hp engineer, network consultant) but noone could track the issue.
Now we tried again and migrated everything, but spantree is still disabled. in the next days we will migrate our firewall(=router) and see what that does with the network.

Another issue we also have on hp 2848/3400's is spanning tree recalculation. I checked every link/log/... and never found a topology change although every half hour(sometimes longer) we have a recalculation.

[Ashley Cooke]

Did anyone ever solve this old case? We've just started reconfiguring an old 5308xl switch to route between VLANs and, when client PCs were migrated into the new VLANs, this same problem has occurred. We are using the latest firmware currently available. Would love to hear from anyone that has been able to resolve, or even work around, this issue.

[Matt Hobbs]

Ashley, please start a new thread if you're having issues with E.11.03. The 5300 has been a stable product now for quite some time.

Alternatively you can contact ProCurve support directly - https://my.procurve.com/help/help_questions.aspx?l2id=48

[cenk sasmaztin]

hi Rasmus
please send me all switch show tech command print .