vitalif
/
vitastor
5
4
Fork 9
Code Issues 1 Pull Requests 2 Actions Releases 68 Wiki Activity
vitastor/docs/config/pool.en.md

13 KiB
Raw Permalink Blame History

Documentation → Configuration → Pool configuration

Читать на русском

Pool configuration

Pool configuration is set in etcd key /vitastor/config/pools in the following JSON format:

{
  "<Numeric ID>": {
    "name": "<name>",
    ...other parameters...
  }
}

Pool configuration is also affected by:

Parameters:

Examples:

Placement Tree

OSD placement tree is set in a separate etcd key /vitastor/config/node_placement in the following JSON format:

{ "<node name or OSD number>": { "level": "<level>", "parent": "<parent node name, if any>" }, ... }

Here, if a node name is a number then it is assumed to refer to an OSD. Level of the OSD is always "osd" and cannot be overriden. You may only override parent node of the OSD which is its host by default.

Non-numeric node names refer to other placement tree nodes like hosts, racks, datacenters and so on.

Hosts of all OSDs are auto-created in the tree with level "host" and name equal to the host name reported by a corresponding OSD. You can refer to them without adding them to this JSON tree manually.

Level may be "host", "osd" or refer to some other placement tree level from placement_levels.

Parent node reference is required for intermediate tree nodes.

OSD settings

Separate OSD settings are set in etc keys /vitastor/config/osd/<number> in JSON format {"<key>":<value>}.

As of now, the following settings are supported:

reweight

  • Type: number, between 0 and 1
  • Default: 1

Every OSD receives PGs proportional to its size. Reweight is a multiplier for OSD size used during PG distribution.

This means an OSD configured with reweight lower than 1 receives less PGs than it normally would. An OSD with reweight = 0 won't store any data. You can set reweight to 0 to trigger rebalance and remove all data from an OSD.

tags

  • Type: string or array of strings

Sets tag or multiple tags for this OSD. Tags can be used to group OSDs into subsets and then use a specific subset for pool instead of all OSDs. For example you can mark SSD OSDs with tag "ssd" and HDD OSDs with "hdd" and such tags will work as device classes.

noout

  • Type: boolean
  • Default: false

If set to true, osd_out_time is ignored for this OSD and it's never removed from data distribution by the monitor.

Pool parameters

name

  • Type: string
  • Required

Pool name.

scheme

  • Type: string
  • Required
  • One of: "replicated", "xor", "ec" or "jerasure"

Redundancy scheme used for data in this pool. "jerasure" is an alias for "ec", both use Reed-Solomon-Vandermonde codes based on ISA-L or jerasure libraries. Fast ISA-L based implementation is used automatically when it's available, slower jerasure version is used otherwise.

pg_size

  • Type: integer
  • Required

Total number of disks for PGs of this pool - i.e., number of replicas for replicated pools and number of data plus parity disks for EC/XOR pools.

parity_chunks

  • Type: integer

Number of parity chunks for EC/XOR pools. For such pools, data will be lost if you lose more than parity_chunks disks at once, so this parameter can be equally described as FTT (number of failures to tolerate).

Required for EC/XOR pools, ignored for replicated pools.

pg_minsize

  • Type: integer
  • Required

Number of available live disks for PGs of this pool to remain active. That is, if it becomes impossible to place PG data on at least (pg_minsize) OSDs, PG is deactivated for both read and write. So you know that a fresh write always goes to at least (pg_minsize) OSDs (disks).

For example, the difference between pg_minsize 2 and 1 in a 3-way replicated pool (pg_size=3) is:

  • If 2 hosts go down with pg_minsize=2, the pool becomes inactive and remains inactive for osd_out_time (10 minutes). After this timeout, the monitor selects replacement hosts/OSDs and the pool comes up and starts to heal. Therefore, if you don't have replacement OSDs, i.e. if you only have 3 hosts with OSDs and 2 of them are down, the pool remains inactive until you add or return at least 1 host (or change failure_domain to "osd").
  • If 2 hosts go down with pg_minsize=1, the pool only experiences a short I/O pause until the monitor notices that OSDs are down (5-10 seconds with the default etcd_report_interval). After this pause, I/O resumes, but new data is temporarily written in only 1 copy. Then, after osd_out_time, the monitor also selects replacement OSDs and the pool starts to heal.

So, pg_minsize regulates the number of failures that a pool can tolerate without temporary downtime for osd_out_time, but at a cost of slightly reduced storage reliability.

FIXME: pg_minsize behaviour may be changed in the future to only make PGs read-only instead of deactivating them.

pg_count

  • Type: integer
  • Required

Number of PGs for this pool. The value should be big enough for the monitor / LP solver to be able to optimize data placement.

"Enough" is usually around 10-100 PGs per OSD, i.e. you set pg_count for pool to (total OSD count * 10 / pg_size). You can round it to the closest power of 2, because it makes it easier to reduce or increase PG count later by dividing or multiplying it by 2.

In Vitastor, PGs are ephemeral, so you can change pool PG count anytime just by overwriting pool configuration in etcd. Amount of the data affected by rebalance will be smaller if the new PG count is a multiple of the old PG count or vice versa.

failure_domain

  • Type: string
  • Default: host

Failure domain specification. Must be "host" or "osd" or refer to one of the placement tree levels, defined in placement_levels.

Two replicas, or two parts in case of EC/XOR, of the same block of data are never put on OSDs in the same failure domain (for example, on the same host). So failure domain specifies the unit which failure you are protecting yourself from.

level_placement

  • Type: string

Additional failure domain rules, applied in conjuction with failure_domain. Must be specified in the following form:

<placement level>=<sequence of characters>, <level2>=<sequence2>, ...

Sequence should be exactly pg_size character long. Each character corresponds to an OSD in the PG of this pool. Equal characters mean that corresponding items of the PG should be placed into the same placement tree item at this level. Different characters mean that items should be placed into different items.

For example, if you want a EC 4+2 pool and you want every 2 chunks to be stored in its own datacenter and you also want each chunk to be stored on a different host, you should set level_placement to dc=112233 host=123456.

Or you can set level_placement to dc=112233 and leave failure_domain empty, because host is the default failure_domain and it will be applied anyway.

Without this rule, it may happen that 3 chunks will be stored on OSDs in the same datacenter, and the data will become inaccessibly if that datacenter goes down in this case.

Of course, you should group your hosts into datacenters before applying the rule by setting placement_levels to something like {"dc":90,"host":100,"osd":110} and add DCs to node_placement, like {"dc1":{"level":"dc"},"host1":{"parent":"dc1"},...}.

raw_placement

  • Type: string

Raw PG placement rules, specified in the form of a DSL (domain-specific language). Use only if you really know what you're doing :)

DSL specification:

dsl := item | item ("\n" | ",") items
item := "any" | rules
rules := rule | rule rules
rule := level operator arg
level := /\w+/
operator := "!=" | "=" | ">" | "?="
arg := value | "(" values ")"
values := value | value "," values
value := item_ref | constant_id
item_ref := /\d+/
constant_id := /"([^"]+)"/

"?=" operator means "preferred". I.e. dc ?= "meow" means "prefer datacenter meow for this chunk, but put into another dc if it's unavailable".

Examples:

  • Simple 3 replicas with failure_domain=host: any, host!=1, host!=(1,2)
  • EC 4+2 in 3 DC: any, dc=1 host!=1, dc!=1, dc=3 host!=3, dc!=(1,3), dc=5 host!=5
  • 1 replica in fixed DC + 2 in random DCs: dc?=meow, dc!=1, dc!=(1,2)

max_osd_combinations

  • Type: integer
  • Default: 10000

Vitastor data placement algorithm is based on the LP solver and OSD combinations which are fed to it are generated ramdonly. This parameter specifies the maximum number of combinations to generate when optimising PG placement.

This parameter usually doesn't require to be changed.

block_size

  • Type: integer
  • Default: 131072

Block size for this pool. The value from /vitastor/config/global is used when unspecified. Only OSDs with matching block_size are used for each pool. If you want to further restrict OSDs for the pool, use osd_tags.

Read more about this parameter in Cluster-Wide Disk Layout Parameters.

bitmap_granularity

  • Type: integer
  • Default: 4096

"Sector" size of virtual disks in this pool. The value from /vitastor/config/global is used when unspecified. Similarly to block_size, only OSDs with matching bitmap_granularity are used for each pool.

Read more about this parameter in Cluster-Wide Disk Layout Parameters.

immediate_commit

  • Type: string, one of "all", "small" and "none"
  • Default: none

Immediate commit setting for this pool. The value from /vitastor/config/global is used when unspecified. Similarly to block_size, only OSDs with compatible bitmap_granularity are used for each pool. "Compatible" means that a pool with non-immediate commit will use OSDs with immediate commit enabled, but not vice versa. I.e., pools with "none" use all OSDs, pools with "small" only use OSDs with "all" or "small", and pools with "all" only use OSDs with "all".

Read more about this parameter in Cluster-Wide Disk Layout Parameters.

pg_stripe_size

  • Type: integer
  • Default: 0

Specifies the stripe size for this pool according to which images are split into different PGs. Stripe size can't be smaller than block_size multiplied by (pg_size - parity_chunks) for EC/XOR pools, or 1 for replicated pools, and the same value is used by default.

This means first pg_stripe_size = (block_size * (pg_size-parity_chunks)) bytes of an image go to one PG, next pg_stripe_size bytes go to another PG and so on.

Usually doesn't require to be changed separately from the block size.

root_node

  • Type: string

Specifies the root node of the OSD tree to restrict this pool OSDs to. Referenced root node must exist in /vitastor/config/node_placement.

osd_tags

  • Type: string or array of strings

Specifies OSD tags to restrict this pool to. If multiple tags are specified, only OSDs having all of these tags will be used for this pool.

primary_affinity_tags

  • Type: string or array of strings

Specifies OSD tags to prefer putting primary OSDs in this pool to. Note that for EC/XOR pools Vitastor always prefers to put primary OSD on one of the OSDs containing a data chunk for a PG.

scrub_interval

  • Type: time interval (number + unit s/m/h/d/M/y)

Automatic scrubbing interval for this pool. Overrides global scrub_interval setting.

used_for_fs

  • Type: string

If non-empty, the pool is marked as used for VitastorFS with metadata stored in block image (regular Vitastor volume) named as the value of this pool parameter.

When a pool is marked as used for VitastorFS, regular block volume creation in it is disabled (vitastor-cli refuses to create images without --force) to protect the user from block volume and FS file ID collisions and data loss.

vitastor-nfs, in its turn, refuses to use pools not marked for the corresponding FS when starting. This also implies that you can use one pool only for one VitastorFS.

The second thing that is disabled for VitastorFS pools is reporting per-inode space usage statistics in etcd because a FS pool may store a very large number of files and statistics for them all would take a lot of space in etcd.

Examples

Replicated pool

{
  "1": {
    "name":"testpool",
    "scheme":"replicated",
    "pg_size":2,
    "pg_minsize":1,
    "pg_count":256,
    "failure_domain":"host"
  }
}

Erasure-coded pool

{
  "2": {
    "name":"ecpool",
    "scheme":"ec",
    "pg_size":3,
    "parity_chunks":1,
    "pg_minsize":2,
    "pg_count":256,
    "failure_domain":"host"
  }
}