diff --git a/howto/ganeti.md b/howto/ganeti.md index ff45ec4db0f78ea374ce3f99e2a6875a56c29937..315e08358b61a2befddd8fc249d7e684185abc67 100644 --- a/howto/ganeti.md +++ b/howto/ganeti.md @@ -2585,11 +2585,11 @@ be reclaimed eventually. `30.172.in-addr.arpa` zone and `torproject.org` zone, in the `admin/dns/domains.git` repository - 4. add the private IP address to the eth1 interface, for example in + 4. add the private IP address to the `eth1` interface, for example in `/etc/network/interfaces.d/eth1`: - auto eth2 - iface eth2 inet static + auto eth1 + iface eth1 inet static address 172.30.131.101/24 Again, this IP must be allocated in the reverse DNS zone file @@ -2598,7 +2598,7 @@ be reclaimed eventually. 5. enable the interface: - ifup eth2 + ifup eth1 6. setup a bridge on the public interface, replacing the `eth0` blocks with something like: diff --git a/howto/quintex.md b/howto/quintex.md index 55b827f50120d7b30c29d47de8c3e7c2088fef1f..84e51f5b31775313ab3da834760ca42724266291 100644 --- a/howto/quintex.md +++ b/howto/quintex.md @@ -514,12 +514,11 @@ This network is split in those VLANs: * "public": VLAN 82 - 204.8.99.0/24, directly accessible on the global network, behind a Quintex router, `eth0` on all nodes, could - eventually be aggregated with `eth1` + eventually be aggregated with `eth2` * "storage": VLAN 801 - 172.30.131.0/24, used by the Ganeti cluster - for DRBD replication, not accessible by the internet, `eth2` on all - nodes (except `dal-rescue-01` where it is `eth1`), could eventually - be aggregated with `eth3` + for DRBD replication, not accessible by the internet, `eth1` on all + nodes, could eventually be aggregated with `eth3` * "OOB": VLAN 802 - 172.30.141.0/24, access to the "out of band" (OOB) management interfaces, not accessible by the internet, @@ -530,6 +529,13 @@ Note that the above use the non-"predictable" interface names, i.e. `eth0` and `eth1` instead of `eno1np0` and `eno1np1` or `enp129s0f0` and `enp129s0f1`. +Also note that have the public and storage VLANs on the *same* NIC +(i.e. public on `eth0` and storage on `eth1`). This is because we plan +on doing aggregation in the long term and that will allow us to +survive a NIC failure. Assuming NIC one has `eth0` and `eth1` and NIC +two has `eth2` and `eth3`, if the public VLAN is on `eth0` and `eth2`, +it will survive a failure of one NIC. + It physically looks like this:  diff --git a/howto/quintex/topology.dot b/howto/quintex/topology.dot index e797a7fbe3fe5f0369b2836494b72ca0dc1974a2..76f2bf637468a91981625e6a0305b5810e054fd2 100644 --- a/howto/quintex/topology.dot +++ b/howto/quintex/topology.dot @@ -4,8 +4,8 @@ digraph quintex { subgraph clusterhosts { label = "server\ndal-node-NN" labeljust=r - host_public [ shape=box label="eth0,eth1\npublic"] - host_storage [ shape=box label="eth2,eth3\nstorage"] + host_public [ shape=box label="eth0,eth2\npublic"] + host_storage [ shape=box label="eth1,eth3\nstorage"] //host_mgmt [ shape=box label="eth2\nmanagement"] host_idrac [ shape=box label="IPMI\nOOB"] } diff --git a/howto/quintex/topology.png b/howto/quintex/topology.png index 623cb0b47af87948ad01de1700bdcf8333c5cf7b..307cd8d9c1b73cf14f16c6719103a556bdf952e1 100644 Binary files a/howto/quintex/topology.png and b/howto/quintex/topology.png differ