Deploy Hyper-Converged Ceph Cluster

From Proxmox VE
Revision as of 12:07, 16 May 2018 by Api (talk | contribs)
Jump to: navigation, search

screenshot/gui-ceph-status.png

Proxmox VE unifies your compute and storage systems, i.e. you can use the same physical nodes within a cluster for both computing (processing VMs and containers) and replicated storage. The traditional silos of compute and storage resources can be wrapped up into a single hyper-converged appliance. Separate storage networks (SANs) and connections via network attached storages (NAS) disappear. With the integration of Ceph, an open source software-defined storage platform, Proxmox VE has the ability to run and manage Ceph storage directly on the hypervisor nodes.

Ceph is a distributed object store and file system designed to provide excellent performance, reliability and scalability.

Some advantages of Ceph on Proxmox VE are:
  • Easy setup and management with CLI and GUI support

  • Thin provisioning

  • Snapshots support

  • Self healing

  • Scalable to the exabyte level

  • Setup pools with different performance and redundancy characteristics

  • Data is replicated, making it fault tolerant

  • Runs on economical commodity hardware

  • No need for hardware RAID controllers

  • Open source

For small to mid sized deployments, it is possible to install a Ceph server for RADOS Block Devices (RBD) directly on your Proxmox VE cluster nodes, see Ceph RADOS Block Devices (RBD). Recent hardware has plenty of CPU power and RAM, so running storage services and VMs on the same node is possible.

To simplify management, we provide pveceph - a tool to install and manage Ceph services on Proxmox VE nodes.

Ceph consists of a couple of Daemons
[Ceph intro https://docs.ceph.com/docs/nautilus/start/intro/]
, for use as a RBD storage:
  • Ceph Monitor (ceph-mon)

  • Ceph Manager (ceph-mgr)

  • Ceph OSD (ceph-osd; Object Storage Daemon)

Tip We highly recommend to get familiar with Ceph’s architecture
[Ceph architecture https://docs.ceph.com/docs/nautilus/architecture/]
and vocabulary
[Ceph glossary https://docs.ceph.com/docs/nautilus/glossary]
.

Precondition

To build a hyper-converged Proxmox + Ceph Cluster there should be at least three (preferably) identical servers for the setup.

Check also the recommendations from Ceph’s website.

CPU

Higher CPU core frequency reduce latency and should be preferred. As a simple rule of thumb, you should assign a CPU core (or thread) to each Ceph service to provide enough resources for stable and durable Ceph performance.

Memory

Especially in a hyper-converged setup, the memory consumption needs to be carefully monitored. In addition to the intended workload from virtual machines and containers, Ceph needs enough memory available to provide excellent and stable performance.

As a rule of thumb, for roughly 1 TiB of data, 1 GiB of memory will be used by an OSD. Especially during recovery, rebalancing or backfilling.

The daemon itself will use additional memory. The Bluestore backend of the daemon requires by default 3-5 GiB of memory (adjustable). In contrast, the legacy Filestore backend uses the OS page cache and the memory consumption is generally related to PGs of an OSD daemon.

Network

We recommend a network bandwidth of at least 10 GbE or more, which is used exclusively for Ceph. A meshed network setup
[Full Mesh Network for Ceph https://pve.proxmox.com/wiki/Full_Mesh_Network_for_Ceph_Server]
is also an option if there are no 10 GbE switches available.

The volume of traffic, especially during recovery, will interfere with other services on the same network and may even break the Proxmox VE cluster stack.

Further, estimate your bandwidth needs. While one HDD might not saturate a 1 Gb link, multiple HDD OSDs per node can, and modern NVMe SSDs will even saturate 10 Gbps of bandwidth quickly. Deploying a network capable of even more bandwidth will ensure that it isn’t your bottleneck and won’t be anytime soon, 25, 40 or even 100 GBps are possible.

Disks

When planning the size of your Ceph cluster, it is important to take the recovery time into consideration. Especially with small clusters, the recovery might take long. It is recommended that you use SSDs instead of HDDs in small setups to reduce recovery time, minimizing the likelihood of a subsequent failure event during recovery.

In general SSDs will provide more IOPs than spinning disks. This fact and the higher cost may make a class based separation of pools appealing. Another possibility to speedup OSDs is to use a faster disk as journal or DB/Write-Ahead-Log device, see creating Ceph OSDs. If a faster disk is used for multiple OSDs, a proper balance between OSD and WAL / DB (or journal) disk must be selected, otherwise the faster disk becomes the bottleneck for all linked OSDs.

Aside from the disk type, Ceph best performs with an even sized and distributed amount of disks per node. For example, 4 x 500 GB disks with in each node is better than a mixed setup with a single 1 TB and three 250 GB disk.

One also need to balance OSD count and single OSD capacity. More capacity allows to increase storage density, but it also means that a single OSD failure forces ceph to recover more data at once.

Avoid RAID

As Ceph handles data object redundancy and multiple parallel writes to disks (OSDs) on its own, using a RAID controller normally doesn’t improve performance or availability. On the contrary, Ceph is designed to handle whole disks on it’s own, without any abstraction in between. RAID controller are not designed for the Ceph use case and may complicate things and sometimes even reduce performance, as their write and caching algorithms may interfere with the ones from Ceph.

Warning Avoid RAID controller, use host bus adapter (HBA) instead.
Note Above recommendations should be seen as a rough guidance for choosing hardware. Therefore, it is still essential to adapt it to your specific needs, test your setup and monitor health and performance continuously.

Initial Ceph installation & configuration

screenshot/gui-node-ceph-install.png

With Proxmox VE you have the benefit of an easy to use installation wizard for Ceph. Click on one of your cluster nodes and navigate to the Ceph section in the menu tree. If Ceph is not already installed you will be offered to do so now.

The wizard is divided into different sections, where each needs to be finished successfully in order to use Ceph. After starting the installation the wizard will download and install all required packages from Proxmox VE’s ceph repository.

After finishing the first step, you will need to create a configuration. This step is only needed once per cluster, as this configuration is distributed automatically to all remaining cluster members through Proxmox VE’s clustered configuration file system (pmxcfs).

The configuration step includes the following settings:

  • Public Network: You should setup a dedicated network for Ceph, this setting is required. Separating your Ceph traffic is highly recommended, because it could lead to troubles with other latency dependent services, e.g., cluster communication may decrease Ceph’s performance, if not done.

  • Cluster Network: As an optional step you can go even further and separate the OSD replication & heartbeat traffic as well. This will relieve the public network and could lead to significant performance improvements especially in big clusters.

screenshot/gui-node-ceph-install-wizard-step2.png

You have two more options which are considered advanced and therefore should only changed if you are an expert.

  • Number of replicas: Defines the how often a object is replicated

  • Minimum replicas: Defines the minimum number of required replicas for I/O to be marked as complete.

Additionally you need to choose your first monitor node, this is required.

That’s it, you should see a success page as the last step with further instructions on how to go on. You are now prepared to start using Ceph, even though you will need to create additional monitors, create some OSDs and at least one pool.

The rest of this chapter will guide you on how to get the most out of your Proxmox VE based Ceph setup, this will include aforementioned and more like CephFS which is a very handy addition to your new Ceph cluster.

Installation of Ceph Packages

Use Proxmox VE Ceph installation wizard (recommended) or run the following command on each node:

pveceph install

This sets up an apt package repository in /etc/apt/sources.list.d/ceph.list and installs the required software.

Create initial Ceph configuration

screenshot/gui-ceph-config.png

Use the Proxmox VE Ceph installation wizard (recommended) or run the following command on one node:

pveceph init --network 10.10.10.0/24

This creates an initial configuration at /etc/pve/ceph.conf with a dedicated network for ceph. That file is automatically distributed to all Proxmox VE nodes by using pmxcfs. The command also creates a symbolic link from /etc/ceph/ceph.conf pointing to that file. So you can simply run Ceph commands without the need to specify a configuration file.

Ceph Monitor

The Ceph Monitor (MON)
[Ceph Monitor https://docs.ceph.com/docs/nautilus/start/intro/]
maintains a master copy of the cluster map. For high availability you need to have at least 3 monitors. One monitor will already be installed if you used the installation wizard. You won’t need more than 3 monitors as long as your cluster is small to midsize, only really large clusters will need more than that.

Create Monitors

screenshot/gui-ceph-monitor.png

On each node where you want to place a monitor (three monitors are recommended), create it by using the Ceph → Monitor tab in the GUI or run.

pveceph mon create

Destroy Monitors

To remove a Ceph Monitor via the GUI first select a node in the tree view and go to the Ceph → Monitor panel. Select the MON and click the Destroy button.

To remove a Ceph Monitor via the CLI first connect to the node on which the MON is running. Then execute the following command:

pveceph mon destroy
Note At least three Monitors are needed for quorum.

Ceph Manager

The Manager daemon runs alongside the monitors. It provides an interface to monitor the cluster. Since the Ceph luminous release at least one ceph-mgr
[Ceph Manager https://docs.ceph.com/docs/nautilus/mgr/]
daemon is required.

Create Manager

Multiple Managers can be installed, but at any time only one Manager is active.

pveceph mgr create
Note It is recommended to install the Ceph Manager on the monitor nodes. For high availability install more then one manager.

Destroy Manager

To remove a Ceph Manager via the GUI first select a node in the tree view and go to the Ceph → Monitor panel. Select the Manager and click the Destroy button.

To remove a Ceph Monitor via the CLI first connect to the node on which the Manager is running. Then execute the following command:

pveceph mgr destroy
Note A Ceph cluster can function without a Manager, but certain functions like the cluster status or usage require a running Manager.

Ceph OSDs

Ceph Object Storage Daemons are storing objects for Ceph over the network. It is recommended to use one OSD per physical disk.

Note By default an object is 4 MiB in size.

Create OSDs

screenshot/gui-ceph-osd-status.png

via GUI or via CLI as follows:

pveceph osd create /dev/sd[X]
Tip We recommend a Ceph cluster size, starting with 12 OSDs, distributed evenly among your, at least three nodes (4 OSDs on each node).

If the disk was used before (eg. ZFS/RAID/OSD), to remove partition table, boot sector and any OSD leftover the following command should be sufficient.

ceph-volume lvm zap /dev/sd[X] --destroy
Warning The above command will destroy data on the disk!
Ceph Bluestore

Starting with the Ceph Kraken release, a new Ceph OSD storage type was introduced, the so called Bluestore
[Ceph Bluestore https://ceph.com/community/new-luminous-bluestore/]
. This is the default when creating OSDs since Ceph Luminous.

pveceph osd create /dev/sd[X]
Block.db and block.wal

If you want to use a separate DB/WAL device for your OSDs, you can specify it through the -db_dev and -wal_dev options. The WAL is placed with the DB, if not specified separately.

pveceph osd create /dev/sd[X] -db_dev /dev/sd[Y] -wal_dev /dev/sd[Z]

You can directly choose the size for those with the -db_size and -wal_size parameters respectively. If they are not given the following values (in order) will be used:

  • bluestore_block_{db,wal}_size from ceph configuration…

    • … database, section osd

    • … database, section global

    • … file, section osd

    • … file, section global

  • 10% (DB)/1% (WAL) of OSD size

Note The DB stores BlueStore’s internal metadata and the WAL is BlueStore’s internal journal or write-ahead log. It is recommended to use a fast SSD or NVRAM for better performance.
Ceph Filestore

Before Ceph Luminous, Filestore was used as default storage type for Ceph OSDs. Starting with Ceph Nautilus, Proxmox VE does not support creating such OSDs with pveceph anymore. If you still want to create filestore OSDs, use ceph-volume directly.

ceph-volume lvm create --filestore --data /dev/sd[X] --journal /dev/sd[Y]

Destroy OSDs

To remove an OSD via the GUI first select a Proxmox VE node in the tree view and go to the Ceph → OSD panel. Select the OSD to destroy. Next click the OUT button. Once the OSD status changed from in to out click the STOP button. As soon as the status changed from up to down select Destroy from the More drop-down menu.

To remove an OSD via the CLI run the following commands.

ceph osd out <ID>
systemctl stop ceph-osd@<ID>.service
Note The first command instructs Ceph not to include the OSD in the data distribution. The second command stops the OSD service. Until this time, no data is lost.

The following command destroys the OSD. Specify the -cleanup option to additionally destroy the partition table.

pveceph osd destroy <ID>
Warning The above command will destroy data on the disk!

Ceph Pools

A pool is a logical group for storing objects. It holds Placement Groups (PG, pg_num), a collection of objects.

Create Pools

screenshot/gui-ceph-pools.png

When no options are given, we set a default of 128 PGs, a size of 3 replicas and a min_size of 2 replicas for serving objects in a degraded state.

Note The default number of PGs works for 2-5 disks. Ceph throws a HEALTH_WARNING if you have too few or too many PGs in your cluster.

It is advised to calculate the PG number depending on your setup, you can find the formula and the PG calculator
[PG calculator https://ceph.com/pgcalc/]
online. From Ceph Nautilus onwards it is possible to increase and decrease the number of PGs later on
[Placement Groups https://docs.ceph.com/docs/nautilus/rados/operations/placement-groups/]
.

You can create pools through command line or on the GUI on each PVE host under Ceph → Pools.

pveceph pool create <name>

If you would like to automatically also get a storage definition for your pool, mark the checkbox "Add storages" in the GUI or use the command line option --add_storages at pool creation.

Further information on Ceph pool handling can be found in the Ceph pool operation
[Ceph pool operation https://docs.ceph.com/docs/nautilus/rados/operations/pools/]
manual.

Destroy Pools

To destroy a pool via the GUI select a node in the tree view and go to the Ceph → Pools panel. Select the pool to destroy and click the Destroy button. To confirm the destruction of the pool you need to enter the pool name.

Run the following command to destroy a pool. Specify the -remove_storages to also remove the associated storage.

pveceph pool destroy <name>
Note Deleting the data of a pool is a background task and can take some time. You will notice that the data usage in the cluster is decreasing.

Ceph CRUSH & device classes

The foundation of Ceph is its algorithm, Controlled Replication Under Scalable Hashing (CRUSH
[CRUSH https://ceph.com/wp-content/uploads/2016/08/weil-crush-sc06.pdf]
).

CRUSH calculates where to store to and retrieve data from, this has the advantage that no central index service is needed. CRUSH works with a map of OSDs, buckets (device locations) and rulesets (data replication) for pools.

Note Further information can be found in the Ceph documentation, under the section CRUSH map
[CRUSH map https://docs.ceph.com/docs/nautilus/rados/operations/crush-map/]
.

This map can be altered to reflect different replication hierarchies. The object replicas can be separated (eg. failure domains), while maintaining the desired distribution.

A common use case is to use different classes of disks for different Ceph pools. For this reason, Ceph introduced the device classes with luminous, to accommodate the need for easy ruleset generation.

The device classes can be seen in the ceph osd tree output. These classes represent their own root bucket, which can be seen with the below command.

ceph osd crush tree --show-shadow

Example output form the above command:

ID  CLASS WEIGHT  TYPE NAME
-16  nvme 2.18307 root default~nvme
-13  nvme 0.72769     host sumi1~nvme
 12  nvme 0.72769         osd.12
-14  nvme 0.72769     host sumi2~nvme
 13  nvme 0.72769         osd.13
-15  nvme 0.72769     host sumi3~nvme
 14  nvme 0.72769         osd.14
 -1       7.70544 root default
 -3       2.56848     host sumi1
 12  nvme 0.72769         osd.12
 -5       2.56848     host sumi2
 13  nvme 0.72769         osd.13
 -7       2.56848     host sumi3
 14  nvme 0.72769         osd.14

To let a pool distribute its objects only on a specific device class, you need to create a ruleset with the specific class first.

ceph osd crush rule create-replicated <rule-name> <root> <failure-domain> <class>

<rule-name>

name of the rule, to connect with a pool (seen in GUI & CLI)

<root>

which crush root it should belong to (default ceph root "default")

<failure-domain>

at which failure-domain the objects should be distributed (usually host)

<class>

what type of OSD backing store to use (eg. nvme, ssd, hdd)

Once the rule is in the CRUSH map, you can tell a pool to use the ruleset.

ceph osd pool set <pool-name> crush_rule <rule-name>
Tip If the pool already contains objects, all of these have to be moved accordingly. Depending on your setup this may introduce a big performance hit on your cluster. As an alternative, you can create a new pool and move disks separately.

Ceph Client

screenshot/gui-ceph-log.png

You can then configure Proxmox VE to use such pools to store VM or Container images. Simply use the GUI too add a new RBD storage (see section Ceph RADOS Block Devices (RBD)).

You also need to copy the keyring to a predefined location for an external Ceph cluster. If Ceph is installed on the Proxmox nodes itself, then this will be done automatically.

Note The file name needs to be <storage_id> + `.keyring - <storage_id> is the expression after rbd: in /etc/pve/storage.cfg which is my-ceph-storage in the following example:
mkdir /etc/pve/priv/ceph
cp /etc/ceph/ceph.client.admin.keyring /etc/pve/priv/ceph/my-ceph-storage.keyring

CephFS

Ceph provides also a filesystem running on top of the same object storage as RADOS block devices do. A Metadata Server (MDS) is used to map the RADOS backed objects to files and directories, allowing to provide a POSIX-compliant replicated filesystem. This allows one to have a clustered highly available shared filesystem in an easy way if ceph is already used. Its Metadata Servers guarantee that files get balanced out over the whole Ceph cluster, this way even high load will not overload a single host, which can be an issue with traditional shared filesystem approaches, like NFS, for example.

screenshot/gui-node-ceph-cephfs-panel.png

Proxmox VE supports both, using an existing CephFS as storage to save backups, ISO files or container templates and creating a hyper-converged CephFS itself.

Metadata Server (MDS)

CephFS needs at least one Metadata Server to be configured and running to be able to work. One can simply create one through the Proxmox VE web GUI’s Node -> CephFS panel or on the command line with:

pveceph mds create

Multiple metadata servers can be created in a cluster. But with the default settings only one can be active at any time. If an MDS, or its node, becomes unresponsive (or crashes), another standby MDS will get promoted to active. One can speed up the hand-over between the active and a standby MDS up by using the hotstandby parameter option on create, or if you have already created it you may set/add:

mds standby replay = true

in the ceph.conf respective MDS section. With this enabled, this specific MDS will always poll the active one, so that it can take over faster as it is in a warm state. But naturally, the active polling will cause some additional performance impact on your system and active MDS.

Multiple Active MDS

Since Luminous (12.2.x) you can also have multiple active metadata servers running, but this is normally only useful for a high count on parallel clients, as else the MDS seldom is the bottleneck. If you want to set this up please refer to the ceph documentation.
[Configuring multiple active MDS daemons https://docs.ceph.com/docs/nautilus/cephfs/multimds/]

Create CephFS

With Proxmox VE’s CephFS integration into you can create a CephFS easily over the Web GUI, the CLI or an external API interface. Some prerequisites are required for this to work:

Prerequisites for a successful CephFS setup:

After this got all checked and done you can simply create a CephFS through either the Web GUI’s Node -> CephFS panel or the command line tool pveceph, for example with:

pveceph fs create --pg_num 128 --add-storage

This creates a CephFS named ‘'cephfs'' using a pool for its data named `'cephfs_data'' with `128` placement groups and a pool for its metadata named 'cephfs_metadata'' with one quarter of the data pools placement groups (`32). Check the Proxmox VE managed Ceph pool chapter or visit the Ceph documentation for more information regarding a fitting placement group number (pg_num) for your setup
[Ceph Placement Groups https://docs.ceph.com/docs/nautilus/rados/operations/placement-groups/]
. Additionally, the `--add-storage’ parameter will add the CephFS to the Proxmox VE storage configuration after it was created successfully.

Destroy CephFS

Warning Destroying a CephFS will render all its data unusable, this cannot be undone!

If you really want to destroy an existing CephFS you first need to stop, or destroy, all metadata servers (M̀DS). You can destroy them either over the Web GUI or the command line interface, with:

pveceph mds destroy NAME

on each Proxmox VE node hosting a MDS daemon.

Then, you can remove (destroy) CephFS by issuing a:

ceph fs rm NAME --yes-i-really-mean-it

on a single node hosting Ceph. After this you may want to remove the created data and metadata pools, this can be done either over the Web GUI or the CLI with:

pveceph pool destroy NAME

Ceph maintenance

Replace OSDs

One of the common maintenance tasks in Ceph is to replace a disk of an OSD. If a disk is already in a failed state, then you can go ahead and run through the steps in Destroy OSDs. Ceph will recreate those copies on the remaining OSDs if possible. This rebalancing will start as soon as an OSD failure is detected or an OSD was actively stopped.

Note With the default size/min_size (3/2) of a pool, recovery only starts when ‘size + 1` nodes are available. The reason for this is that the Ceph object balancer CRUSH defaults to a full node as `failure domain’.

To replace a still functioning disk, on the GUI go through the steps in Destroy OSDs. The only addition is to wait until the cluster shows HEALTH_OK before stopping the OSD to destroy it.

On the command line use the following commands.

ceph osd out osd.<id>

You can check with the command below if the OSD can be safely removed.

ceph osd safe-to-destroy osd.<id>

Once the above check tells you that it is save to remove the OSD, you can continue with following commands.

systemctl stop ceph-osd@<id>.service
pveceph osd destroy <id>

Replace the old disk with the new one and use the same procedure as described in Create OSDs.

Trim/Discard

It is a good measure to run fstrim (discard) regularly on VMs or containers. This releases data blocks that the filesystem isn’t using anymore. It reduces data usage and resource load. Most modern operating systems issue such discard commands to their disks regularly. You only need to ensure that the Virtual Machines enable the disk discard option.

Scrub & Deep Scrub

Ceph ensures data integrity by scrubbing placement groups. Ceph checks every object in a PG for its health. There are two forms of Scrubbing, daily cheap metadata checks and weekly deep data checks. The weekly deep scrub reads the objects and uses checksums to ensure data integrity. If a running scrub interferes with business (performance) needs, you can adjust the time when scrubs
[Ceph scrubbing https://docs.ceph.com/docs/nautilus/rados/configuration/osd-config-ref/#scrubbing]
are executed.

Ceph monitoring and troubleshooting

A good start is to continuosly monitor the ceph health from the start of initial deployment. Either through the ceph tools itself, but also by accessing the status through the Proxmox VE API.

The following ceph commands below can be used to see if the cluster is healthy (HEALTH_OK), if there are warnings (HEALTH_WARN), or even errors (HEALTH_ERR). If the cluster is in an unhealthy state the status commands below will also give you an overview of the current events and actions to take.

# single time output
pve# ceph -s
# continuously output status changes (press CTRL+C to stop)
pve# ceph -w

To get a more detailed view, every ceph service has a log file under /var/log/ceph/ and if there is not enough detail, the log level can be adjusted
[Ceph log and debugging https://docs.ceph.com/docs/nautilus/rados/troubleshooting/log-and-debug/]
.

You can find more information about troubleshooting
[Ceph troubleshooting https://docs.ceph.com/docs/nautilus/rados/troubleshooting/]
a Ceph cluster on the official website.