Netgear GS724Tv3 - odd behaviour

[ddaenen1]

I have a strange situation with my managed Netgear GS724Tv3 switch and/or Mikrotik RB3011 router. Everything seems to be running flawless and then at some point in time, i cannot access the switch anymore. At the same time, i can see in my DHCP server list that the switch did not get a new IP lease. Oddly enough, everything seems to keep on running fine so all network and internet connections just continue. I just cannot access the switch anymore and the switch cannot access the web anymore for NTP updates and stuff.

I have not tried factory resetting either of them as this will require everything to be being down for some time which is not easy in a house full of life.

I have noticed though that the switch did not get a clientID in the router DHCP server whilst all other devices do get one. I also know this has started when i tried to give the switch a custom client ID in the router which made it go offline. I did cold start the switch after that to get access to it again and have done it again a second time but i can't see to get the DHCP server to forget the switch and assign it to a new DHCP lease after i did a cold reset.

Any recommendations s where to look first or options to find the root cause without having the factory reset both router and switch?

 

[coxhaus]

So you tried both IP addresses, the old IP and the new clientID IP? Can you delete the static DHCP entry in the Mikrotik? Then maybe stop and start DHCP.
I am not a Mikrotik guy but I seen this kind of problem on other devices.

 

[ddaenen1]

So you tried both IP addresses, the old IP and the new clientID IP? Can you delete the static DHCP entry in the Mikrotik? Then maybe stop and start DHCP.
I am not a Mikrotik guy but I seen this kind of problem on other devices.

I am still unsure what the issue is but i have a suspicion the problem is the switch, not the router as it appears to be the case that the machine that makes a DHCP request includes the client ID which in this case is the Netgear switch.

As i cannot seem to allocate the reall root cause, i did a full factory reset on both the router and switch and configured them from scratch. Now it is just a matter of monitoring if the switch keeps its IP.

 

[ColinTaylor]

Are you using the latest firmware for your switch? I see that there was a DHCP fix a few releases back:

Bug Fixes:

• Fixed the issue of switch not able to obtain IP address from some DHCP server. This particular DHCP server requires the source address from the DHCP discovery packet to be 0.0.0.0.

 

[coxhaus]

Sounds like it does not work with DHCP reservations. So to me it also sounds like it should of still worked with the original IP address.

 

[sfx2000]

I have a strange situation with my managed Netgear GS724Tv3 switch and/or Mikrotik RB3011 router. Everything seems to be running flawless and then at some point in time, i cannot access the switch anymore. At the same time, i can see in my DHCP server list that the switch did not get a new IP lease. Oddly enough, everything seems to keep on running fine so all network and internet connections just continue. I just cannot access the switch anymore and the switch cannot access the web anymore for NTP updates and stuff.

For this very reason, I've found it best practice for core elements to not use DHCP, but rather fix them with static IP's outside of the DHCP scope.

This would include things like AP's, L2/L3 managed switches, and Servers - DHCP is very useful for client devices...

 

[coxhaus]

This would include things like AP's, L2/L3 managed switches, and Servers - DHCP is very useful for client devices...

Yes separate IPs from DHCP works best for network devices in large networks. The only issue is you need to maintain a manual document with the list of IP to what they are. Maybe a small spread sheet.

 

[ColinTaylor]

The only issue is you need to maintain a manual document with the list of IP to what they are. Maybe a small spread sheet.

Typically you would still want to have the host names of these devices registered in DNS. So whatever system you use to create those entries is effectively the document.

 

[sfx2000]

I do (and did) both - on the enterprise/carrier scale, network elements were in DNS, and a managed text flat file over on our SVN (and then GIT) internal server was maintained.

 

[coxhaus]

Typically you would still want to have the host names of these devices registered in DNS. So whatever system you use to create those entries is effectively the document.

We never bothered with DNS for our Cisco gear. We had thousands of PCs on this network spread across more than 50 sites.

 

[sfx2000]

Typically you would still want to have the host names of these devices registered in DNS. So whatever system you use to create those entries is effectively the document.

See above, and having the flat file is a backstop - one person managed both...

 

[sfx2000]

We never bothered with DNS for our Cisco gear. We had thousands of PCs on this network spread across more than 50 sites.

Yep, like I mentioned, DHCP is great for clients, but core/essential elements should be static - everything can be in DNS, and a good DNS package (infoblox DNS is what we used as a server cluster - and that includes a few million devices on our customer network)

 

[coxhaus]

We were not telco we were a private network. We used AT&T for telecommunications. I never saw a need for host names on Cisco gear. We had standards for IP addresses for routers and switches. We did everything by telnet and IP. It seemed to work fine.

 

[ColinTaylor]

See above, and having the flat file is a backstop - one person managed both...

Looks like we're all using variations on the same basic idea . Similar situation here; enterprise setup across hundreds of sites, everything on clustered servers but with one very large spreadsheet as the backstop. The hardest part is making sure 20+ infrastructure guys (and gals) remember to keep it up to date and accurate every time they make a change to the network.

 

[sfx2000]

Similar situation here; enterprise setup across hundreds of sites, everything on clustered servers but with one very large spreadsheet as the backstop. The hardest part is making sure 20+ infrastructure guys (and gals) remember to keep it up to date and accurate every time they make a change to the network.

True, true...

We had the "DNS Boss" that kept things aligned... we had four (at a high level) networks - internal for the business, the engineering plant, the customer network (handsets), and the public facing network for the company website and support for our end-users...

2 corp HQ's (SAN/DEN), 3 data centers (DEN/IAD/ORD), 10,000 cell sites, 7 million+ devices, and of course the front end to the customer for support and paying their bill online.

Lot's of private blocks for IPv4 internally, we started deploying CGN in a market basis (100.64.0.0/10) as part of the IPv6 transition... I know folks hate the CGN, but it's actually a step forward to IPv6...

$ ipcalc 100.64.0.0/10
Address:   100.64.0.0           01100100.01 000000.00000000.00000000
Netmask:   255.192.0.0 = 10     11111111.11 000000.00000000.00000000
Wildcard:  0.63.255.255         00000000.00 111111.11111111.11111111
=>
Network:   100.64.0.0/10        01100100.01 000000.00000000.00000000
HostMin:   100.64.0.1           01100100.01 000000.00000000.00000001
HostMax:   100.127.255.254      01100100.01 111111.11111111.11111110
Broadcast: 100.127.255.255      01100100.01 111111.11111111.11111111
Hosts/Net: 4194302               Class A

Flat Text Files are actually better than an excel worksheet - most of us are in linux/unix land, so parsing that file is very fast and easy.

 

[ddaenen1]

Are you using the latest firmware for your switch? I see that there was a DHCP fix a few releases back:

Yes. I am running the latest available from their website.

 

[ddaenen1]

Sounds like it does not work with DHCP reservations. So to me it also sounds like it should of still worked with the original IP address.

Well, the full factory reset for both router and switch did not fix the problem as only a couple of hours later, same thing and i also know it happens when the router refreshes the lease. Yesterday, as an interim containment, i have configured the DHCP lease time for the switch to 30 days. Since then everything seems to be working fine. I will let this run for a couple of weeks and see if the connection remains stable. If so, I am going to give it a go with a static address though as a permanent fix.

 

[ddaenen1]

For this very reason, I've found it best practice for core elements to not use DHCP, but rather fix them with static IP's outside of the DHCP scope.

This would include things like AP's, L2/L3 managed switches, and Servers - DHCP is very useful for client devices...

Well, i have read different recommendations on this in the forum. I have everything inside the DHCP range and wired devices such as the switch, file server, NAS, AP's all have static mappings in DHCP. I am going to see what the 30 day lease time does for the switch. After that, i might try out giving it a static IP.

 

[ddaenen1]

Well, i have read different recommendations on this in the forum. I have everything inside the DHCP range and wired devices such as the switch, file server, NAS, AP's all have static mappings in DHCP. I am going to see what the 30 day lease time does for the switch. After that, i might try out giving it a static IP.

Seems that the switch doesn't like the DHCP server from the Mikrotik router. Since i changed the lease time to 30 days, everything has been running flawless for 6 days now. Let's see if that stays the same for the remainder of the 30 day lease.