I have a RB 5009 running 7.11 that has on ether2-ether8 various hardware attached. These are servers with a management board - not an operating system that I install, so my influence is limited on what to do with them. In the past, they did get their IP from a windows OS based virtual machine running also DHCP, but I am retiring this mechanism and moving towards a full mikrotik implementation on this layer. No need for a heavy VM just to do some internal DNS and DHCP for what essentially is a hardware management network.
I just added the DHCP server and of the 4 currently attached machines, I could see 4 leases appearing, 2 of them offered, to bound. THen the offered disappeared after some time. I restart the ethernet ports that are not bound (disable/enable) and I get another one bound, the last one is quite resistant in accepting the offer.
Anyone has any logical explanation for that outside of the “bug”? There is no VLAN involved yet, the port configurations are identical. I replaced cables - no change, so either it is a quite arcane issue in the ports (which work fine otherwise) or a bug in the mikrotik dhcp or other firmware elements?
Before anyone says " it is a DHCP configuration issue" - again: 2 of 4 machines work, another one after a restart of the port. All the same DHCP config. So, no - that is not the issue.
Would be helpful if you can share your config.
/export file=anynameyoulike
Don’t forget to remove the serial and any other private information (like public IP address).
On your VLAN issue you also initially insisted it was a bug.
And it turned out to be a config issue.
So it could still be a config issue (I am not entirely ruling out a bug but the forum would be swamped with error reports if that truly was the case).
Export of latest config is already a start.
Also check log file for DHCP related messages, specifically those related to MAC address of the bad boys who are not listening.
You can add debug messages (temporarily): go to system logging and select debug, prefix dhcp
Personal comment:
I always assign fixed IP to any server when possible (and mark that address as static in IP pool). I don’t want a server to become unresponsive because of a failing DHCP service.
But that’s me.
The problem i see is that this can just not be a config issue. I also did not insist the VLAN issue was a BUG - I asked for config advice.
It works or it does not. It has addresses or it has not.
The other end is a firmware - not an installed operating system. They are obviously set to ask for IP (because they all do that).The port configs are identical for the ports, they all are on the same untagged network, and I see the requests in. The DHCP server obviously is working (as it is handing out IP) and it is not out of addresses - I see the IP addresses being offered, just not binding.
I simply do not see where the config can be wrong here. If it would be a DHCP config issue, DHCP would not work - not “not work for SOME hardware but good for the majority”. And again, the pools are not exhausted obviouly.
Also note - your idea of assigning static IP for a server is semi-valid. Semi because, you know, you do not get hardware BMC’s with a static IP so at least once you must do it dynamically and two, that is “just” the BMC. I am slowly thinking about retiring the whole IPV4 line on that layer anyway (and IPv6 does not have the DHCP intrinsic issues of v4, mostly that you must have the DHCP server running when the link goes up or nothing happens later). What would move them to a static IP anyway. These literally are BMC’s - I use them to check hardware status when I need something, or mostly reinstall the OS once every other year.
dhcpconfig.rsc (7.39 KB)
Long shot after quickly looking at your config:
I’ve seen in the past DHCP clients not getting a DHCP lease if the time settings between client and server are off.
On Cisco ISP modems acting as DHCP server.
It can’t hurt to check:
I see in your DHCP settings you specify the router as NTP server.
I don’t see IP Cloud, NTP Client, nor DHCP client.
Where does it get it’s time from then ?
That is a good question - I would assume they have some sort of uplink because both machines have not only the correct time within the same second (as per system/cloud), but also show the correctly identified time zone (by auto detect).
What is funny is that you do not see it - must be because it does not dump default values?
Per Winbox all my Mikrotik routers are defined to pull time from Cloud - Update Time is set to enabled. That is the time source. I assume as I never changed it - it runs on default values, and thus is not included in the configuration?
That’s correct. Default values are not being shown in export.
But then we can strike that possibility out already.
Hm. I just ordered a new set of higher level patch cables. Those look awfully loose to me actually. See what Amazon delivers tomorrow.
Fixed. Crappy cableS - plural. I replaced thhem, did not work - the plugs are too loose. Tomorrow I get high end plugs, for now it just works. AAAAAH
Point is, it is not “a cable”. I got 10 of them and I started switching them yesterday around and the same problems on pretty much ALL of them.
Yes, definitely COVID - highly contagious. I got it working now by slapping them in way behind the point where the guard snaps into place. Well… tomorrow I get new ones. CAT 8 - overkill, but physically super well made.
