[doc-updates] doc: be slightly more critical of LMDB

This commit is contained in:
Alex 2024-03-14 16:29:42 +01:00
parent 238545e564
commit a7dddebedd
Signed by untrusted user: lx
GPG key ID: 0E496D15096376BE
3 changed files with 53 additions and 31 deletions

View file

@ -27,7 +27,7 @@ To run a real-world deployment, make sure the following conditions are met:
[Yggdrasil](https://yggdrasil-network.github.io/) are approaches to consider
in addition to building out your own VPN tunneling.
- This guide will assume you are using Docker containers to deploy Garage on each node.
- This guide will assume you are using Docker containers to deploy Garage on each node.
Garage can also be run independently, for instance as a [Systemd service](@/documentation/cookbook/systemd.md).
You can also use an orchestrator such as Nomad or Kubernetes to automatically manage
Docker containers on a fleet of nodes.
@ -53,9 +53,9 @@ to store 2 TB of data in total.
### Best practices
- If you have fast dedicated networking between all your nodes, and are planing to store
very large files, bump the `block_size` configuration parameter to 10 MB
(`block_size = 10485760`).
- If you have reasonably fast networking between all your nodes, and are planing to store
mostly large files, bump the `block_size` configuration parameter to 10 MB
(`block_size = "10M"`).
- Garage stores its files in two locations: it uses a metadata directory to store frequently-accessed
small metadata items, and a data directory to store data blocks of uploaded objects.
@ -68,20 +68,29 @@ to store 2 TB of data in total.
EXT4 is not recommended as it has more strict limitations on the number of inodes,
which might cause issues with Garage when large numbers of objects are stored.
- If you only have an HDD and no SSD, it's fine to put your metadata alongside the data
on the same drive. Having lots of RAM for your kernel to cache the metadata will
help a lot with performance. Make sure to use the LMDB database engine,
instead of Sled, which suffers from quite bad performance degradation on HDDs.
Sled is still the default for legacy reasons, but is not recommended anymore.
- For the metadata storage, Garage does not do checksumming and integrity
verification on its own. If you are afraid of bitrot/data corruption,
put your metadata directory on a ZFS or BTRFS partition. Otherwise, just use regular
EXT4 or XFS.
- Servers with multiple HDDs are supported natively by Garage without resorting
to RAID, see [our dedicated documentation page](@/documentation/operations/multi-hdd.md).
- For the metadata storage, Garage does not do checksumming and integrity
verification on its own. Users have reported that when using the LMDB
database engine (the default), database files have a tendency of becoming
corrupted after an unclean shutdown (e.g. a power outage), so you should use
a robust filesystem such as BTRFS or ZFS for the metadata partition, and take
regular snapshots so that you can restore to a recent known-good state in
case of an incident. If you cannot do so, you might want to switch to Sqlite
which is more robust.
- LMDB is the fastest and most tested database engine, but it has the following
weaknesses: 1/ data files are not architecture-independent, you cannot simply
move a Garage metadata directory between nodes running different architectures,
and 2/ LMDB is not suited for 32-bit platforms. Sqlite is a viable alternative
if any of these are of concern.
- If you only have an HDD and no SSD, it's fine to put your metadata alongside
the data on the same drive, but then consider your filesystem choice wisely
(see above). Having lots of RAM for your kernel to cache the metadata will
help a lot with performance.
## Get a Docker image
Our docker image is currently named `dxflrs/garage` and is stored on the [Docker Hub](https://hub.docker.com/r/dxflrs/garage/tags?page=1&ordering=last_updated).
@ -187,7 +196,7 @@ upgrades. With the containerized setup proposed here, the upgrade process
will require stopping and removing the existing container, and re-creating it
with the upgraded version.
## Controling the daemon
## Controlling the daemon
The `garage` binary has two purposes:
- it acts as a daemon when launched with `garage server`
@ -245,7 +254,7 @@ You can then instruct nodes to connect to one another as follows:
Venus$ garage node connect 563e1ac825ee3323aa441e72c26d1030d6d4414aeb3dd25287c531e7fc2bc95d@[fc00:1::1]:3901
```
You don't nead to instruct all node to connect to all other nodes:
You don't need to instruct all node to connect to all other nodes:
nodes will discover one another transitively.
Now if your run `garage status` on any node, you should have an output that looks as follows:
@ -328,8 +337,8 @@ Given the information above, we will configure our cluster as follow:
```bash
garage layout assign 563e -z par1 -c 1T -t mercury
garage layout assign 86f0 -z par1 -c 2T -t venus
garage layout assign 6814 -z lon1 -c 2T -t earth
garage layout assign 212f -z bru1 -c 1.5T -t mars
garage layout assign 6814 -z lon1 -c 2T -t earth
garage layout assign 212f -z bru1 -c 1.5T -t mars
```
At this point, the changes in the cluster layout have not yet been applied.

View file

@ -57,7 +57,7 @@ to generate unique and private secrets for security reasons:
cat > garage.toml <<EOF
metadata_dir = "/tmp/meta"
data_dir = "/tmp/data"
db_engine = "lmdb"
db_engine = "sqlite"
replication_mode = "none"

View file

@ -264,18 +264,31 @@ Performance characteristics of the different DB engines are as follows:
- Sled: tends to produce large data files and also has performance issues,
especially when the metadata folder is on a traditional HDD and not on SSD.
- LMDB: the recommended database engine on 64-bit systems, much more
space-efficient and slightly faster. Note that the data format of LMDB is not
portable between architectures, so for instance the Garage database of an
x86-64 node cannot be moved to an ARM64 node. Also note that, while LMDB can
technically be used on 32-bit systems, this will limit your node to very
small database sizes due to how LMDB works; it is therefore not recommended.
- LMDB: the recommended database engine for high-performance distributed
clusters, much more space-efficient and significantly faster. LMDB works very
well, but is known to have the following limitations:
- The data format of LMDB is not portable between architectures, so for
instance the Garage database of an x86-64 node cannot be moved to an ARM64
node.
- While LMDB can technically be used on 32-bit systems, this will limit your
node to very small database sizes due to how LMDB works; it is therefore
not recommended.
- Several users have reported corrupted LMDB database files after an unclean
shutdown (e.g. a power outage). This situation can generally be recovered
from if your cluster is geo-replicated (by rebuilding your metadata db from
other nodes), or if you have saved regular snapshots at the filesystem
level.
- Sqlite: Garage supports Sqlite as an alternative storage backend for
metadata, and although it has not been tested as much, it is expected to work
satisfactorily. Since Garage v0.9.0, performance issues have largely been
fixed by allowing for a no-fsync mode (see `metadata_fsync`). Sqlite does not
have the database size limitation of LMDB on 32-bit systems.
metadata, which does not have the issues listed above for LMDB.
On versions 0.8.x and earlier, Sqlite should be avoided due to abysmal
performance, which was fixed with the addition of `metadata_fsync`.
Sqlite is still probably slower than LMDB due to the way we use it,
so it is not the best choice for high-performance storage clusters,
but it should work fine in many cases.
It is possible to convert Garage's metadata directory from one format to another
using the `garage convert-db` command, which should be used as follows:
@ -302,7 +315,7 @@ Using this option reduces the risk of simultaneous metadata corruption on severa
cluster nodes, which could lead to data loss.
If multi-site replication is used, this option is most likely not necessary, as
it is extremely unlikely that two nodes in different locations will have a
it is extremely unlikely that two nodes in different locations will have a
power failure at the exact same time.
(Metadata corruption on a single node is not an issue, the corrupted data file