Omada SDN Controller - SG2428P Switch
I have a couple of posts i want to make about issues in SDN - possibly quick fixes, but this topic is about an issue i can't seem to figure out.
My setup
EAP245V3 (x2) - Connected all under the same WLAN
SG2428P (about 3 days old), configured and connected to Omada SDN - hence the topic being under this category
ISP - PfSense (VM) - ESXi - 2 x LAN ports out (non LAG/LACP) - SG2428P
Omada SDN controller running on Debian, clean install, no migrated data. All FW and OS versions are the latest.
I have posted my question here, though i believe the issue is around configuration of the switch, simply because the switch is managed by Omada, so all guides on the switch configuration itself are not relevant or have not been helpful in me finding a solution.
I have read both the Switch manual and the Omada controller manual, though i'll be honest, i may have missed or misunderstood something.
Do note that all of the above setup was the same previously with the exception of the switch, I've replaced an SG2434P with SG2428P, the configuration otherwise is all the same.
My issues;
1. I can see the switch, all ports and all physically conencted devices, where a name is listed an IP is too and the 'uptime' is stable and counting as it should, however for any VMs on the ESXi host, connected by 2 LAN ports, not LAG/LACP configured, i get a list of mac addresses, but no IP and no host names (I've added some manually) and in the log>events, these connections connect and disconnect constantly, I thought this might be the two NICs, so i disconnected one, the problem persists. Do note, the devices do not physically drop off the network, they do not drop pings either, but they log a disconnection and connection.
The PfSense firewall has multiple VLANs and both physical NICS are trunks (in VMware) and all tagged except LAN on the switch. Using the ALL profile
Nothing disconnects physically and i can always ping VMs, the LAN ports, the host, firewall etc. It's just the disconnection in the events view, it shouldn't be doing this and since i don't drop any pings, i dont know if this is accurate or something i've missed in the switch/Omada configuration. This is affecting all VMs and all VMs will use ports 23 or 24.
2. My EAP 245v3, one of them has a pass-through device connected, I've set the LAN port as a different VLAN (but same as one of the SSIDs the AP hosts) - but it never gives the guest an IP on the VLAN it's marked for, it only ever gives an IP from the VLAN the AP is connected to, if this is VLAN1, the pass-through device gets an IP from VLAN1, if this is VLAN30 it gets an IP from VLAN30, even if it's marked to be on VLAN 50 for example.
3. Also related and similar to 2. If i put the AP on a different VLAN, such as VLAN30, but also want guests on the SSID to use 30, 40 and 50, again as examples, SSIDs 40 and 50 both work, but 30 fails to get an IP - yet the AP is happily sitting on this VLAN. If i move the AP to VLAN1, all 3 SSIDs work as expected. In the settings of the switch, it says the native VLAN is tagged by default, so VLAN30 should be passing through, but it doesn't seem to. This setup worked on the old switch, just doesn't seem to on the new one.
Sorry this is a long post, i hope it all makes sense and i've simply missed something in the configuration - the configuration between my old switch and the Omada managed one is a little different, so it's possible my translation of the configuration isn't 10% accurate.
I will post the issues with Omada SDN seperately another time.
- Copy Link
- Subscribe
- Bookmark
- Report Inappropriate Content
I can accept idle VMs might drop their entries in a table, such as arp cache, but if a ping is continually running, then the entry should still be there (in theory).
This isnt what i am seeing.
Even so, it's not that the entry is there for, 8 hours then drops and reconnects, the time varies, but it's usually short, less than a half hour.
In addition some of the VMs are static IPs others are DHCP - much like they would be in a business - i plan to do soem further testing over the weekend, i have a few spare NIC ports i can light up and use for this should i need to.
I might even enable promiscuous mode on the virtual switch to see if that has any impact.
Just as an example - notice this can be up to about an hour but is usually quite a low value, the traffic of 0 bytes is also something i am debating given this device specificlaly is a DNS server as well as doing DHCP
Do please remember that i am by no means complaining, just hoping someone says something or notices something i can use to aid me - after all i am keen to learn and if we never figure this out, at least we tried
- Copy Link
- Report Inappropriate Content
The more i play around with settings and the more i use the Omada SDN product, the more i think there is something miscalulating.
Just so i am clear, my install is on Debian (Kali), clean install with nothing imported, so it can't be a corruption in an upgrade.
I am seeing devices, both physically connected and virtual machines, connected via 2 ports (23/24) showing me traffic of 0bytes, which i know is not true, even the most idle system will use something.
My backup system - runs as a VM, but shows 0 traffic too, this on a nightly basis will be trasnferring ~200GB of data to an external NAS which must pass the physical switch to work, yet the VM shows me no traffic at all.
As additional information to my original post - sporadically some of my VMs show an IP, but it's rare and doesn't stay visible, i also have 2 maybe 3 VMs which are showing conencted for 10+ days, so i know they can be seen and the switch IS registering the correct mac address.
I am not complaining, but hoping that providing further information is helpful in someway to figuring out what's going on. Is it simply my misunderstanding of networking or is there something amiss
I hope that as more people adopt the SDN and compatible switches, someone more knowledgable on network kit can voice their views too.
- Copy Link
- Report Inappropriate Content
@Rod-IT Hello Rod, I was trying to figure out the same since I joined this system, than I renounced to understand because this is not affecting anything but it is happening the same to me. My ISP router is connected to one wire to the switch TL-SG2428P and just that. In the event log I see exactly what you describe, an interface (which I named TIM HUB Interface) that belongs to the modem/router where I see a MAC address without any IP address that is connecting and disconnecting in variable time frame but multiple times per day. Not an issue but I wanted to understand why.
- Copy Link
- Report Inappropriate Content
Could you also test something for me?
If you do to the device page - does your switch or switches show 0 bytes until you refresh it?
I am running my Omada SDN on Debian and as yet i am not seeing the 4.2.2 release to try it, but there are a lot of little quirks - some the Android app deals with better than the web-browser version. The data is clearly there and is being put in to the DB, but i have a feeling tables are being mixed when something tries to read it - the other day my NAS, according to the software passed 65TB of data - more than i have actual storage, but a refresh changed it back to 6.5TB, I still don't think this is right, but more plausible than 65TB.
The disconnections are not in-fact disconnections and this is not for all devices, but it's a pain seeing it as it messes with the whole concept of visualizing the devices and their stability. Something perhaps we could troubleshoot better if the options the switch has available to it, were usable via the SDN controller instead of being stripped back for simplicity. If linked the switch to SDN reduces it's feature set, then don't include the features in the switch at all - the SDN is the switches USP.
Both SDN and the switches we have are new, so i am willing to give TP time to resolve these issues, however, as you've noted in previous posts, this is somewhat basic functionalities - these things should be ironed out and ready to roll long before they are available for customers to use. Heck, I'd even be a beta tester for new products or software (not so much on my live setups, but generically).
- Copy Link
- Report Inappropriate Content
To confirm, the other devices which are connected to the SG2428P switch don't have such problem, right?
In my view, it seems that the issue is related to the uplink device(non-Omada Router/Gateway) connected to the SG2428P switch, the Clients page would not show the IP address of the uplink device and the event log keeps showing the uplink device is disconnected with 0 traffic but it's actually connected without any problem.
I guess that the SDN Controller is unable to know the IP address and the actual traffic of the front non-Omada Router/Gateway due to some reason.
- Copy Link
- Report Inappropriate Content
Not completely.
The switch is able to get the devices MAC addresses of all systems that are virtual or using uplinks or NICs with multiple mac addresses/VMs and *sometimes* it does show the IP, but not always, when it does show an IP it doesn't match the VLAN name, for example if i seen IP 192.168.10.10 this would be a client network but it might show in Omada as Servers. Though seeing an IP is hit and miss - I'm not overly concerned about this, but it would be nice.
The data transferred of these devices is not always 0bytes, but mostly 0bytes. I do see actual data for the VMs depending on how long they remain connected (according to the switch).
As an example; both of these below are virtual machines and both behind a NIC that has multiple mac addresses/VMs.
My email server shows uptime of 2 hours and no bytes - this is untrue, but it shows what i mean, yet my backup server also virtual shows uptime of less than an hour and 9mb transferred.
Both of these servers are connected to the same switch ports and reside on the same virtual host.
The disconnections also happen with physically devices, though not as often - an example, my NAS, it's a physical box connected to the switch via 2 non-bonded or LACP/LAG ports, just 2 direct connections - this also shows disconnections.
On the other side of this, i have 1 VM that shows connected 31 days and zero disconnections, but it is on the same virtual host and using the same NICs/Ports as all my other VMs, so i'm confused why this one is available and stable, yet the rest are showing disconnects - even though physically they never go offline.
Because this is not causing any actual problems i am not too concerned, but the SDN software is the reason i bought this switch and to see it lose half it's functionalities when managed by Omada or to see issues like this, but for unknown reasons and to suggest it's everything else and not the switch just doesn't make sense.
If i'd have known any of this at the start, prior to purchase, in all honesty i would have likely stayed with my existing switch as this gives me little to no benefits at this point.
The switch does record the device history based on MAC, so it clearly knows of the devices.
I hope this helps and if there is any information i can get from the switch or the software, please let me know and how.
- Copy Link
- Report Inappropriate Content
Could you also test something for me?
If you do to the device page - does your switch or switches show 0 bytes until you refresh it?
No Rod, I see several GB in the UP and DOWN columns, I have no idea if those numbers are true, but definitely I don't see zero. Just for your info I'm using OC200 with version 4.2.4, but I don't remind with software controller or previous 4.1.5 version to have seen zero here.
Regarding the limited functionality of the Omada, it's something already raised up, and I agree with you that we must have the possibilities to use all the functionalities of all devices plus the benefits of the Omada otherwise it loses of sense. But from my understanding this is quite new and with time the things will come, or well I hope so. For example AP are supported since longer time and you have access to almost all functionalities while switches or router are more new into this system and that's why they are in earlier stage I think.
Aswell for the several bugs or things implemented in a questionable way I think it's because the system is pretty new and they will need time to fix and improve.
From our side we can help them to find and highlight things.
@Fae No other devices do not have this behaviour, only the ISP router and I guess because multiple interfaces are in, but I don't understand why I see those MAC interfaces if these are disconnected.
- Copy Link
- Report Inappropriate Content
I just want to inform everyone of something, specifically you and the team.
I today updated to 4.2.4 using the deb package linked in the sticky topics, i imported my configuration and since then i can now see 4 IPs of devices coming from my VMware host, which are using ports 23 and 24 - which was not working prior, so this is additional proof that this is in fact possible and the issue is either software or FW on the switch.
The devices still show as connected and disconnected, but they dont actually disconnect, but given i can now see the IPs of a handful of devices, it must be the way in which the software is talking to the switch. I don't see all IPs, but i see 4 which is 4 more than before.
If the switch was purely L2 i would understand why it doesn't see IPs, but it has some L3 and L4 features which seem to be disabled or redcuced when Omada SDN takes ownership - which should not be the case, but at present it does.
I'd like to regain full control over the switch and the features it has in Omada SDN, but if this was not possible, maybe the switches local administration access could be retained as well - just a though.
- Copy Link
- Report Inappropriate Content
I want to add additional information.
I am sporadically seeing more devices on the up-links and their IPs, but they don't always stay, sometimes the IP appear all the time, others come and go.
I am leaning more towards this being an SDN configuration issue (backend), due to other odd results i am seeing.
For example - my NAS is showing transferred data in the 800TB range which is impossible, I don't think it will have read this amount of data in it's lifetime, this however will fix itself, if i go back and look later this figure will be a more realistic number.
Again, i want to be clear i am not complaining, simply trying to help find and squash these issues
- Copy Link
- Report Inappropriate Content
Information
Helpful: 0
Views: 5601
Replies: 29
Voters 0
No one has voted for it yet.