summaryrefslogtreecommitdiffstats
path: root/doc/rbd/rbd-persistent-write-log-cache.rst
blob: af323962d0c61f8ff3e5836e3e25bc40385dfb57 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
================================
 RBD Persistent Write Log Cache
================================

.. index:: Ceph Block Device; Persistent Write Log Cache

Persistent Write Log Cache
===========================

The Persistent Write Log Cache (PWL) provides a persistent, fault-tolerant
write-back cache for librbd-based RBD clients.

This cache uses a log-ordered write-back design which maintains checkpoints
internally so that writes that get flushed back to the cluster are always
crash consistent. Even if the client cache is lost entirely, the disk image is
still consistent but the data will appear to be stale.

This cache can be used with PMEM or SSD as a cache device. For PMEM, the cache
mode is called ``replica write log (rwl)``. At present, only local cache is
supported, and the replica function is under development. For SSD, the cache
mode is called ``ssd``.

Usage
=====

The PWL cache manages the cache data in a persistent device. It looks for and
creates cache files in a configured directory, and then caches data in the
file.

The PWL cache depends on the exclusive-lock feature. The cache can be loaded
only after the exclusive lock is acquired.

The cache provides two different persistence modes. In persistent-on-write mode,
the writes are completed only when they are persisted to the cache device and
will be readable after a crash. In persistent-on-flush mode, the writes are
completed as soon as it no longer needs the caller's data buffer to complete
the writes, but does not guarantee that writes will be readable after a crash.
The data is persisted to the cache device when a flush request is received.

Initially it defaults to the persistent-on-write mode and it switches to
persistent-on-flush mode after the first flush request is received.

Enable Cache
========================================

To enable the PWL cache, set the following configuration settings::

        rbd_persistent_cache_mode = {cache-mode}
        rbd_plugins = pwl_cache

Value of {cache-mode} can be ``rwl``, ``ssd`` or ``disabled``. By default the
cache is disabled.

The ``rwl`` cache mode depends on libpmem library (part of PMDK). It should
be universally available on x86_64 architecture and may also be available on
ppc64le and aarch64 architectures on some distributions. It is not available
on s390x architecture.

Here are some cache configuration settings:

- ``rbd_persistent_cache_path`` A file folder to cache data. This folder must
  have DAX enabled (see `DAX`_) when using ``rwl`` mode to avoid performance
  degradation.

- ``rbd_persistent_cache_size`` The cache size per image. The minimum cache
  size is 1 GB.

The above configurations can be set per-host, per-pool, per-image etc. Eg, to
set per-host, add the overrides to the appropriate `section`_ in the host's
``ceph.conf`` file. To set per-pool, per-image, etc, please refer to the
``rbd config`` `commands`_.

Cache Status
------------

The PWL cache is enabled when the exclusive lock is acquired,
and it is closed when the exclusive lock is released. To check the cache status,
users may use the command ``rbd status``.  ::

        rbd status {pool-name}/{image-name}

The status of the cache is shown, including present, clean, cache size and the
location as well as some basic metrics.

For example::

        $ rbd status rbd/foo
        Watchers:
                watcher=10.10.0.102:0/1061883624 client.25496 cookie=140338056493088
        Persistent cache state:
                host: sceph9
                path: /mnt/nvme0/rbd-pwl.rbd.101e5824ad9a.pool
                size: 1 GiB
                mode: ssd
                stats_timestamp: Sun Apr 10 13:26:32 2022
                present: true   empty: false    clean: false
                allocated: 509 MiB
                cached: 501 MiB
                dirty: 338 MiB
                free: 515 MiB
                hits_full: 1450 / 61%
                hits_partial: 0 / 0%
                misses: 924
                hit_bytes: 192 MiB / 66%
                miss_bytes: 97 MiB

Flush Cache
-----------

To flush a cache file with ``rbd``, specify the ``persistent-cache flush``
command, the pool name and the image name.  ::

        rbd persistent-cache flush {pool-name}/{image-name}

If the application dies unexpectedly, this command can also be used to flush
the cache back to OSDs.

For example::

        $ rbd persistent-cache flush rbd/foo

Invalidate Cache
----------------

To invalidate (discard) a cache file with ``rbd``, specify the
``persistent-cache invalidate`` command, the pool name and the image name.  ::

        rbd persistent-cache invalidate {pool-name}/{image-name}

The command removes the cache metadata of the corresponding image, disables
the cache feature and deletes the local cache file if it exists.

For example::

        $ rbd persistent-cache invalidate rbd/foo

.. _section: ../../rados/configuration/ceph-conf/#configuration-sections
.. _commands: ../../man/8/rbd#commands
.. _DAX: https://www.kernel.org/doc/Documentation/filesystems/dax.txt