Skip to content

Commit

Permalink
Duplicate hardware requirements in quicktart guide
Browse files Browse the repository at this point in the history
  • Loading branch information
Markus Napp committed Sep 3, 2018
1 parent 5079379 commit a2117bf
Showing 1 changed file with 111 additions and 46 deletions.
157 changes: 111 additions & 46 deletions xml/quick_system_requirements.xml
Original file line number Diff line number Diff line change
Expand Up @@ -26,10 +26,10 @@
<para>
&productname; is a dedicated cluster operating system and only functions in a
multi-node configuration. It requires a connected group of four or more
physical or virtual machines.
physical or virtual machines.
</para>
<para>
The minimum supported cluster size is four nodes: a single &admin_node;, one
The minimum supported cluster size is four nodes: a single &admin_node;, one
&master_node;, and two &worker_node;s.
</para>
<note>
Expand All @@ -42,59 +42,124 @@
<para>
For improved performance, multiple &master_node;s are supported, but there
must always be an odd number. For cluster reliability, when using multiple
master nodes, some form of DNS load-balancing should be used.
master nodes, some form of DNS load-balancing should be used.
</para>
<para>Any number of &worker_node;s may be added up to the maximum cluster
size. For the current maximum supported number of nodes, please refer to
the Release Notes on <link xlink:href="https://www.suse.com/releasenotes/"/>.
</para>
</sect1>
<sect1 xml:id="sec.quick.requirements.hardware">
<title>Minimum Hardware Requirements</title>

<itemizedlist mark="bullet" spacing="normal">
<listitem>
<para>
Each node must have an AMD64/Intel* EM64T processor; quad-core processors
are recommended. 32-bit processors are not supported.
</para>
</listitem>
<listitem>
<para>
40 GB available disk space (or more depending on the count and size of the
container images you intend to run).
</para>
</listitem>
<listitem>
<para>
8 GB physical RAM per node.
</para>
</listitem>
</itemizedlist>
<note>
<title>Swap Partitions</title>
<para>
A requirement of &kube; means that &productname; &productnumber; does
not support swapping to disk. Due to a technical limitation in the
installation program, the &admin_node; will have a small swap partition, but
this is inactive and not used. Nodes installed using AutoYaST have no swap.
</para>
<para>
Thus, if a node has insufficient physical RAM available, its operation may
fail. RAM sizing is thus more important than for typical Linux operating
systems.
</para>
</note>

<sect1 xml:id="sec.quick.requirements.hardware">
<title>Minimum Node Specification</title>
<para>
Each node in the cluster must meet the following minimum specifications. All
these specifications must be adjusted according to the expected load and type
of deployments.
</para>
<variablelist>
<varlistentry>
<term>(v)CPU</term>
<listitem>
<itemizedlist>
<listitem>
<para>
AMD64/Intel* EM64T processor (4 core highly recommended)
</para>
</listitem>
<listitem>
<para>
32-bit processors are not supported
</para>
</listitem>
</itemizedlist>
</listitem>
</varlistentry>
<varlistentry>
<term>Memory</term>
<listitem>
<itemizedlist>
<listitem>
<para>
<literal>8 GB</literal>
</para>
<para>
Although it may be possible to install &productname; with less memory
than recommended, there is a high risk that the operating system will
run out of memory and subsequently causes a cluster failure.
</para>
<note>
<title>Swap partitions</title>
<para>
&kube; does not support swap.
</para>
<para>
For technical reasons, an &admin_node; installed from an ISO image will
have a small swap partition which will be disabled after installation.
Nodes built using &ay; do not have a swap partition.
</para>
</note>
</listitem>
</itemizedlist>
</listitem>
</varlistentry>
<varlistentry>
<term>Storage Size</term>
<listitem>
<itemizedlist>
<listitem>
<para>
<literal>40 GB</literal>
for the root file system with Btrfs and enabled snapshots.
</para>
<note>
<title>Cloud default root volume size</title>
<para>
In some Public Cloud frameworks the default root volume size of the
images is smaller than 40GB. You must resize the root volume before
instance launch using the command line tools or the web interface for
the framework of your choice.
</para>
</note>
</listitem>
</itemizedlist>
</listitem>
</varlistentry>
<varlistentry>
<term>Storage Performance</term>
<listitem>
<itemizedlist>
<listitem>
<para>
IOPS: <literal>500</literal> sequential IOPS
</para>
</listitem>
<listitem>
<para>
Write Performance: <literal>10MB/s</literal>
</para>
<note>
<title>etcd Storage requirements</title>
<para>
Storage performance requirements are tied closely to the
<link xlink:href="https://github.com/etcd-io/etcd/blob/master/Documentation/op-guide/hardware.md#disks">etcd hardware recommendations</link>
</para>
</note>
</listitem>
</itemizedlist>
</listitem>
</varlistentry>
</variablelist>
</sect1>

<sect1 xml:id="sec.quick.requirements.network">
<title>Network Requirements</title>

<itemizedlist mark="bullet" spacing="normal">
<listitem>
<para>
All nodes should to be on the same network. It is important that the nodes
can communicate without interruptions.
can communicate without interruptions.
</para>
</listitem>
<listitem>
Expand All @@ -106,7 +171,7 @@
<listitem>
<para>
It is recommended to let the DHCP server provide not only the IP addresses
but also the host names of the cluster nodes.
but also the host names of the cluster nodes.
</para>
</listitem>
<listitem>
Expand All @@ -124,7 +189,7 @@
</listitem>
<listitem>
<para>
Both these FQDNs are required during installation. They cannot
Both these FQDNs are required during installation. They cannot
subsequently be added or changed.
</para>
</listitem>
Expand All @@ -139,7 +204,7 @@
<para>
All nodes must be able to resolve the name of the &admin_node;. It is not
required for &worker_node;s to have resolvable names, although this is not
problematic and may aid troubleshooting.
problematic and may aid troubleshooting.
</para>
</listitem>
<listitem>
Expand Down Expand Up @@ -175,12 +240,12 @@
host name, bootstrap of the cluster will fail.
</para>
</important>

<note>
<title>Clusters without Fully-Qualified Domain Names</title>
<para>
For test purposes, IP addresses can be substituted for the FQDNs for the
&admin_node; and &master_node;, but this is not supported for production
&admin_node; and &master_node;, but this is not supported for production
deployment.
</para>
</note>
Expand All @@ -199,7 +264,7 @@
<listitem>
-->
<!-- cwickert 2017-04-06: FIXME - check if still applies
https://bugzilla.suse.com/show_bug.cgi?id=1029317
https://bugzilla.suse.com/show_bug.cgi?id=1029317
jahalackova: that's still valid for 1.0.
lproven 2018-03-27: and for 3
-->
Expand Down

0 comments on commit a2117bf

Please sign in to comment.