Text Blame History Raw

Hardware notes

Seamicro Chassis

Seamicro was a brand that vanished from market but that's the hardware we still use for CentOS CI infra. It's composed of :

  • hardware chassis working in "fabric" mode
  • network switch[es]
  • management (inband and out-of-band)
  • storage pool (64 * SSD)
  • compute card (64) each with dedicated cpu/memory and link to network/storage pool

Worth knowing that Seamicro doesn't provide webui for management, nor is ipmi compliant. There is also no vga/keyboard/mouse interaction with the chassis nor the compute nodes (servers).

There are only two ways to interact with chassis :

  • rest api exposed on the management ip of the chassis
  • cli (ssh into management card)

We'll cover here some basic operations to power-on/off a server, take a remote SoL console (each compute node emulates ttyS0 so all installs are using serial redirection, as no vga, for console access).

Just keep in mind that indexing starts at 0 so first compute node is 0/0 and second one 1/0 and so on.

The CLI looks like cisco IOS so if you're familiar with it, it would all make sense.

Note

we keep all chassis configuration in a git repository (that can be reused to reconfigure a whole chassis easily, including network/disks/vlan/etc)

node operations

ssh into admin@<chassis>

See summary/status

show server summary <n>/0/g

Power Off

enable; power-off server <n>/0/g [force]

Power on

enable; power-on server <n>/0/g [force]

reset a node

enable; reset server <n>/0/g

Forcing a boot over pxe

enable ; reset server <n>/0/g using-pxe

Chassis operation

ssh into admin@<chassis>

Power off all nodes

enable; power-off server all [force]

Power on all nodes

enable; power-on server all [force]

Reset chassis

enable; power-off server all [force]
reload

Access serial console of a seamicro node

ssh into admin@<chassis>
enable; server console connect <n>/0/g