The only way I know how to get infiniband (mellanox card) running under YaST is to select the Wicked service and configure the indicated IB port. This works well. But I can’t seem to get any of the server’s ethernet ports (em1 - em4) to work, either with a static IP address or DHCP when using Wicked. (the ethernet is connected to a standard router, and IB to a separate IB switch - they are separate subnets) On the other hand, it’s easy to get em1 to work fine, I just choose Network Manager - but NM can’t seem to set up the IB port correctly.
Now I’ve run into a situation where I need both IB (say, p1p1 and em1) running at the same time. Can someone help me with what I’m missing here?
Hi pattiM. With NM both interfaces appear to be active, but you have no ip address assigned for p1p1. Did you try to configure that interface explicitly via the connection editor?
YaST - Switched to Wicked
=========================
~> ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
<snip!>
4: em3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
link/ether f0:4d:a2:02:d3:6d brd ff:ff:ff:ff:ff:ff
inet 198.162.1.186/24 brd 198.162.1.255 scope global noprefixroute dynamic em3
valid_lft 603817sec preferred_lft 603817sec
<snip!>
6: p1p1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
link/ether xx:xx:xx:xx:xx:xx brd ff:ff:ff:ff:ff:ff
inet xxx.xxx.xxx.xxx/xx brd xxx.xxx.xxx.xxx scope global p1p1
valid_lft forever preferred_lft forever
~>
~> ip route
default via xxx.xxx.xxx.xxx dev em3 proto dhcp metric 101
xxx.xxx.xxx.xxx/xx dev p1p1 proto kernel scope link src xxx.xxx.xxx.xxx
xxx.162.xxx.xxx/xx dev em3 proto kernel scope link src xxx.xxx.xxx.xxx metric 101
~>
Here you appear to have both interfaces configured and working as desired?
Power-cycled
============
~> ip addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
<snip!>
4: em3: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
link/ether xx:xx:xx:xx:xx:xx brd ff:ff:ff:ff:ff:ff
<snip!>
6: p1p1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
link/ether xx:xx:xx:xx:xx:xx brd ff:ff:ff:ff:ff:ff
inet xxx.xxx.xxx.xxx/xxx brd xxx.xxx.xxx.xxx scope global p1p1
valid_lft forever preferred_lft forever
~>
~> ip route
xxx.xxx.xxx.xxx/xxx dev p1p1 proto kernel scope link src xxx.xxx.xxx.xxx
~>
Only p1p1 active? Can you bring the em3 interface up manually with ip command?
I’ll try your command to bring em up manually when using Wicked.
My examples were to show:
Wicked: em doesn’t work but p1p1 does work
NM: the other way around (em* works, but NM can’t can’t find what it needs to complete the set-up of p1p1)
In my perception, Wicked is ‘old-school,’ sort of - I enter IP/SNM statically and it works fine for p1p1 (IB). The IB LAN is just a Mikrotik switch (no router) between some compute nodes, but the switch has a pre-setup IP range, so I assign appropriate IP/SNM to network nodes manually (with Wicked).
(Here’s the switch: https://mikrotik.com/product/crs309_1g_8s_in )
I have no idea what Network Manager is doing behind-the-scenes. There’s only one input I recognize in NM, and that’s DNS, which I have never needed to enter.
(but there’s no DNS on the IB switch LAN)
I tried setting up em* with Wicked using either DHCP or assigned IP address (I got the IP address by letting NM set up em*, then copying the IP address from ifconfig)
I also tried the options for network-start at boot time or on-cable-connection. it seems like that should have worked.
Of course - this morning the computer is hanging at the splash-screen (right after Wicked setup) :beat-up:
But ALL my computers have shown this duality between Wicked and NM - so this non-boot is just a typical computer morning-PITA
NM: the other way around (em* works, but NM can’t can’t find what it needs to complete the set-up of p1p1)*
OK - after the booting thing this AM, I noticed both em* and p1p1 were up and working using Wicked.
I had set Wicked to DHCP and IP4 (…did I never do that combination before? It seems so straightforward)
Still need to test, but “apparently” a reboot was all that was needed? (even with the option to activate-upon-cable-connect).
side question: Is IP6 needed for reliable Internet data-connections these days? (e.g., downloading binary data from NOAA weather-data servers)
Thanks for the feedback, I guess I get stuck on stuff sometimes if I can’t get feedback, so thank you very much.
I can’t believe I wasted your guyz time with something that should have been so simple and straightforward. :embarrassed:rotfl!
I’m testing repeatedly right now - but it looks solved with Wicked (simple/robust is good).
Geeze, EVERYTHING has been hacked, hasn’t it? Sakari, Accelion, Spectre, ESS Voting Machines, MS Exchange, Signed PDFs, SolarWinds, BrowserExtensions…
…makes me wonder how long the *nix kernels will be safe!