Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[WillieC]

I recently got a new DGS-1100-08 to add to the existing 2 DGS-1210-10 and 1 oter DGS-1100-08 switches in my network.  The configuration I'm trying to accomplish looks like this:
https://1drv.ms/i/s!AhB_FiZe3glrpQoervvIQvYvN0aT

The circles labeled 1, 2, 3 and 4 are LAGs.  1 and 2 are LACP. 3 and 4 are static because that's all the 1100 series will do.. 1, 2 and 3 have been there and working fine for a while.  Then I added 4 and when PC1 reboots or wakes up from sleep I get a multicast broadcast storm that takes the whole network down.

Here's a screen shot of what's happening shown in Wireshark:
https://1drv.ms/i/s!AhB_FiZe3glrpQkaS5iGaRVRhCTJ

PC1 (10.0.0.101) is the source of this storm. If I pull one of the cables from EITHER link 3 OR 4 the storm stops.  If I reconfigure either 3 or 4 without the port trunking the storm doesn't happen. If I don't reconfigure and just leave the cable unplugged the storm doesn't happen. I've tried swapping out the switches and it doesn't change so it's not a defective switch.  It seems there is a loop happening somewhere and it has to do with link aggregation between the 2 DGS-1100-08s.  As far as I can tell this shouldn't happen of if it does it should happen whether port trunking is enabled or not.

Help! Any ideas?

[FurryNutz]

Seems like either a cabling port issue or a configuration issue possibly. If you go back to your old configuration with out #4 it all works correctly?
Just curious, why so many managed switches? 

[WillieC]

Yeah, it works fine without #4.  I've replaced all the cables as well and it doesn't help.

The switches are kind of a least cost aquisition thing. Bought an 1100-08  and it didn't quite fill my needs so bought a 1210 now I've got 2 etc.  Also there are VLANs on all this stuff that need managed switched.  What I've shown here is all on the same VLAN and same subnet so wasn't really relevant to the issue.

[FurryNutz]

Seems like a possible configuration issue here or something with #4 or Port Trunking?
I'll get some eyes on this. PacketTracer is really good with switches and such.

[PacketTracer]

Hi,

it looks like both DGS-1100-08 do not have a proper LAG configuration, hence they do not treat the links that comprise the LAG as a single logical link (with a multiple of the bandwidth of a single link) but as individual links. This means that any ethernet frame received from one of these "LAG-wanted" links, that is a broadcast or a multicast or that is destined for an unknown MAC address is flooded to any other port (this is standard switch behaviour), unfortunately including the other "LAG-wanted" links (this must not happen, if the LAG were configured properly).

As long as only one DGS-1100-08 is connected, no loop is formed, because DGS-1210-10 seems to have a proper (local) configuration of its LAG end and would not send  broadcast/multicast/unknown frames received from one LAG link back down another LAG link. Only if the second DGS-1100-08 is connected, a loop is formed between both  DGS-1100-08, because the second DGS-1100-08 reveals the same misconfiguration and hence provides the "frame reflector" (this way "closing" the loop) that was so far missing in your scenario with a single DGS-1100-08.

I assume STP is switched off? You could (should!) switch it on. This would help to prevent loops, especially those that happen by accidentially connecting switch ports that shouldn't be connected. STP would also break your present loop by blocking all "LAG wanted" links except one in each "wanted LAG". While good for switch sanity this is not what you want, because this results in singly linked DGS-1100-08 switches - equivalent to having no LAGs (#3, #4) configured at all.

Hence, check your LAG configuration: Did you configure it the same way at both ends (DGS-1100-08 <-- LAG --> DGS-1210-10) of the LAG ("static")? Are all links that comprise the LAG configured the same way, e.g. for the same VLAN or set of VLANs (if VLAN trunk)? This is what the manual says to "port trunk groups" (LAGs):

"All of the ports in the group must be members of the same VLAN, and their STP status, static multicast, traffic control; traffic segmentation and 802.1p default priority configurations must be identical."

PT

[WillieC]

Thanks for the response. Sounds like I'm kind of stuck due to the 1100 limitations.

When I got the second 1100-08 I loaded the configuratiuon from the first one and then modified it for the different VLANs that it supports. Other than those changes the configurations are identical. No STP (for the reasons you state), no trafic control, no traffic segmentation.  I did play with storm control on all 3 switches but it wasn't effective so I turned it back off again.

Here are some screenshots of the 1100-08 on 3. (4 is currently configured without the LAG)
https://1drv.ms/i/s!AhB_FiZe3glrpQ1R1w19LhZ6t8k-
https://1drv.ms/i/s!AhB_FiZe3glrpQ4hckOl9pClY7VX

And here are the same configuration items on the 1210-10 The LAG on this switch that's not configured would be on ports 5 & 6.
https://1drv.ms/i/s!AhB_FiZe3glrpQyl6btTdwhGnPlJ
https://1drv.ms/i/s!AhB_FiZe3glrpQvMRiTIIyagAm-w

Alternately I could always swap out one of the 1100-08s with the other 1210-10 I've got that doesn't show in these diagrams (connected to port 9 on the 1210-10) that switch doesn't really need any functionality that the 1100 can't supply. No LAGs etc. Just a royal PITA to swap them.

[PacketTracer]

... missing the links following "... here are ..."

[WillieC]

I re-pasted the links. Try it now.

[FurryNutz]

 Adding Screenshots In A Post

[WillieC]

Just what I did. problem is I don't see the results only the [img] tags with the link.  SO I can't really tell if it worked or not.

[FurryNutz]

Your links are set for HTTP so use the 2nd icon over to link to HTTP content. Your links are HTTP URL based.
[img]https://1drv.ms/i/s!AhB_FiZe3glrpQ4hckOl9pClY7VX/img to [url]https://1drv.ms/i/s!AhB_FiZe3glrpQ4hckOl9pClY7VX/url
Brackets of course at the ends.

Photobucket uses this kind of string to get pictures here in the forums:
URL=http://s1195.photobucket.com/user/furrynutz740il/media/989_zps5590a04d.jpg.html][IMGhttp://i1195.photobucket.com/albums/aa396/furrynutz740il/989_zps5590a04d.jpg[/img[/URL

[WillieC]

Thanks again for the help.  I changed the previous post to the [url] tags.  Next time I'l look into what you wrote about getting the actual image to show up.

Again thanks!

[PacketTracer]

Hi again,

looking at your screenshots nothing looks wrong: For LAG #3 ports 1 and 2 (port trunking group 1) at DGS-1100-08 are configured to be connected to ports 7 and 8 (port trunking group 3) at DGS-1210-10 and to transmit tagged frames for VLANs 100 and 120. Of course, the physical cabling could not match this configuration, but I presume you have carefully checked, that ports 1,2 at DGS-1100-08 are connected to ports 7,8 at DGS-1210-10.

VLAN configuration also requires PVIDs to be specified for all ports of a switch. What does the PVID configuration look like for the relevant ports that comprise LAG #3? PVID is for ingess untagged frames, which should not happen at the LAG ports (since only frames tagged for VLANs 100 and 120 should be received), but nevertheless the switches require a PVID value to be configured. I propose you consistently use PVID=1 (as a "dummy" value) for all LAG ports (I wouldn't use a PVID value for a port that matches a tagged VID configured for that port).

Btw: Do you use the latest firmware versions (for the US: 1.10.039 for DGS-1110-08-REVA and 4.10.004 for DGS-1210-10-REVC)?

PT

[WillieC]

I really appreciate your time on this.  As is often the case with this kind of thing, in talking to you  I came up with a solution to my immediate problem by switching the one 1100-08 with another 1210-10. Interestingly with 2 1210s and 1 1100-08 in the mix the multicast storm still starts when I wake the PC from sleep but it calms down and goes back to normal in a few seconds. Since Wireshark goes to sleep and and stops recording when the PC does I can't show you a picture of this.

I obviously can't show you a meaningfull picture of the actual wiring but here is the diagram I work off of when doing it:
https://1drv.ms/i/s!AhB_FiZe3glrpRO8zoAEQdWDknJs

This (and all the rest of these) are already modified with the switch swap I did last night. (Change the docs first then build from the docs.) Since the changes are made I'm going to show you the configs on the other 1100-08 marked 4 in the net diagram and labeled Asylum-Switch-03 on the switch config drawing. With the exception of the VLANs enabled it is the same as the other 1100 was.

This is the PVID setting for the 1210-10. It's as you suggested.
https://1drv.ms/i/s!AhB_FiZe3glrpRLOn41NyzIy-sVT

And here for the 1100. Since the management VLANs on these are set to 100 it won't let me set the PVID on the LAG to 1. Works fine on the 1210 but barks on the 1100. The 1210 automatically sets the PVIDs on trunks to 1 and the 1100 automatically sets them to the management VLAN VID.
https://1drv.ms/i/s!AhB_FiZe3glrpQ-7PgzY4RyWi5Fe

Here are the system settings for both switches (1210 first). Firmware is all up to date.
https://1drv.ms/i/s!AhB_FiZe3glrpRGkypdArTnZeUFd
https://1drv.ms/i/s!AhB_FiZe3glrpRBshlCV3Myjhctm

Interesting factoid here. I bought the 2 1210s over a month apart yet the serial numbers are only seperated by 1.

Once again I really appreciate your time on this. It's been a big help.  It once again confirms my opinion that the 1100 series switches are an excersize in frustration. They just don't quite do whatever you want them to.  Mush better to spend the extra and get a 1210 series instead.

[PacketTracer]

Hi again,

And here for the 1100. Since the management VLANs on these are set to 100 it won't let me set the PVID on the LAG to 1. Works fine on the 1210 but barks on the 1100. The 1210 automatically sets the PVIDs on trunks to 1 and the 1100 automatically sets them to the management VLAN VID.

Can you overwrite the PVID-Settings for the LAG ports at DGS-1110-08 with PVID=1 if you declare those ports to be also members of VLAN 1 (untagged) in a preceding step (while maintaining their membership in VLANs 100 and 120, respectively)?

PT