-
Notifications
You must be signed in to change notification settings - Fork 37
-
Notifications
You must be signed in to change notification settings - Fork 37
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
SNMP-less servers are no longer connected in the NAV topology #1753
Comments
This is related to topology. There are some discussions regarding this in the archive on the nav-users mailing list. Do a search for "topology". The answer for Navtopology Difficulties and VLAN Bandwidth load I would guess is relevant. |
no luck so far to identify the issue. |
@a31amit Please clarify: Am I correct in understanding that the ipdevinfo page shows uplink information for a given server, yet it is isolated in the layer 2 netmap view? |
@lunkwill42 let me know if you needed more details. just to add a little more note for my configuration. that I am using latest 4.8.5 Appliance and I have modified cron frequency which is running with navcron user. I try to delete, re-create devices but that didn't help me. |
Right, @a31amit , and what are your actual view settings in Netmap when this happens? |
Hi, is there any update for this? I have the same problem, running 4.8.6. I am quite confident my netmap was right while on 4.7.x but it has been quite some time since I took a look. Funny thing is, some servers DO so connected, but most are not. But even these are missing their second interface connection to another switch. Even my gateway shows isolated! For example, direct neighborship candidates report show this for ceph-10-206-13-178 and host-10-206-123-174:
And this is my netmap: As you can see, of the above, only ceph-10-206-123-178 show connected and only to one switch, not the other. I've changed the dell5324 and n016 switches to GSW from SW to allow ip2mac to run (it used to run for SW too but stopped after an upgrade on 17-4-2018, maybe that's when netmap broke too...) in hopes that the situation is changed somehow but it didnt. |
@koukou73gr , if I understand you correctly, ceph-10-206-123-178.physics.auth.gr should appear as connected to n016.physics.auth.gr, but it doesn't. You provided this piece of collected candidate data:
But: Do you have candidate data from ceph-10-206-123-178.physics.auth.gr that corroborates this? As part of the changed topology algorithm in NAV 4.8, LLDP data that cannot be corroborated by both peers will be thrown out by NAV. Also, thanks for providing ample data - it makes the bug report a lot more useful. |
Nope! Nothing. No reverse entry. Just the switch to server connection is there, not vice versa. There is no reverse even for the connection of ceph-10-206-123-178 to dell5324 switch that IS drawn on the netmap!
Whatever else I can collect and present you, please ask. If there is some short of debug switch to increase output, please say so. |
@koukou73gr Well, that would explain why |
No. And it never did. None of my servers ever did as I am now finding out it needs more configuration in each server. I never had this, but still, once (upon a time) my netmap was correct. |
Right, reconfiguring lldpd on the server in question to connect as an snmp subagent did the trick. Is this now a requirement? I don't recall reading anything in 4.8.x release notes. |
that doesn't seem to be true in my case, I can see a few servers are showing connected but direct neighbors report their source as "CAM" rather LLDP [ that means cam was used to generate netmap ] and some are not. I can see them connected in Device Info and Ports view. Sw - Arista over SNMP |
@koukou73gr The topology detector now requires LLDP neighbor reports to be bilateral, since NAV 4.8 (we had a lot of issue with HP devices that reflected partial CDP information into LLDP records, which caused topolical chaos). I can see this was not explicitly stated in the release notes, and in the case of servers, it might not have been fully thought through. It was never a requirement that servers needed SNMP support at all, so we may need to rethink this decision. |
@lunkwill42 , thanks for confirming this. And sorry for hijacking the issue, it appears @a31amit has a slightly different problem. |
@a31amit I agree with koukou73gr that you two are probably not having the same issue, so I ask again: What are your actual view settings in Netmap when the problem occurs? |
Hm, @a31amit, your example from above shows a device with an uplink to I see a similar issue on an installation I have access to, and although the switch that shows as isolated in Netmap has a known uplink, the topology information on the other end of that link does not include the port information - which is probably why Netmap seems to discard it. |
So we have moved that particular server, But I can confirm from another server which has case information that I connection doesn't show any port. only device name/ip infornation along with VLAN detail is mentioned. |
@a31amit , it is definitely because NAV does not have port information from the server. It could only get that if the server either has an SNMP agent, or has an LLDP agent (in which case, the port would be identified when pulling the LLDP records from the switch). Nevertheless, your NAV has the data that is needed to draw a proper link in Netmap, so I would say the bug is likely in Netmap, and is likely caused by Netmap ignoring the link because of the missing neighboring port information. |
When the source has multiple distinct neighbor candidates, we cannot trust a random one, so only trust when there is a single distinct candidate. Closes #1753
I have some servers without snmp and switches with snmp. However, netmap weather map doesn't show some of them connected.
Those servers are showing uplink to switches and ethernet port information is visible however it's not showing them connected in netmap weather map and show them as isolated nodes.
How can I troubleshoot it? please help
The text was updated successfully, but these errors were encountered: