From: Deepika <dupadhya@redhat.com>
Date: Fri, 19 Mar 2021 14:34:51 +0000 (+0000)
Subject: doc/dev: define what mempools we use in bluestore
X-Git-Tag: v18.0.0~1064^2~1
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=d6b0dc3049eb52384351bd5b789149f1cbef4f62;p=ceph-ci.git

doc/dev: define what mempools we use in bluestore

Signed-off-by: Deepika <dupadhya@redhat.com>
---

diff --git a/doc/dev/mempool_accounting.rst b/doc/dev/mempool_accounting.rst
new file mode 100644
index 00000000000..1ec8fd6d790
--- /dev/null
+++ b/doc/dev/mempool_accounting.rst
@@ -0,0 +1,56 @@
+What is a mempool?
+------------------
+A memory pool is a method for accounting the consumption of memory of
+a set of containers.
+Each memory pool tracks the number of bytes and items it contains.
+Memory pools are statically declared and serves the purpose of identifying
+memory related checks in BlueStore.
+
+Some common mempools that we can track
+--------------------------------------
+
+- ``bloom_filter``: Ceph uses bloom filters to account for tracking objects already
+  in the cache pool, in order to determine which objects in the pool are being
+  accessed.
+- ``bluestore_alloc``: It accounts for actual allocations done by an allocator.
+- ``bluestore_inline_bl``: measures encoded length of an inline buffer.
+- ``bluestore_fsck``: file system consistency check for BlueStore metadata, helps in
+  tracking and debugging during OSD repairs.
+- ``bluestore_txc``: accounts for committed transactions counter.
+- ``bluestore_writing_deferred``: measures those small writes which are deferred
+  i.e. first written into RocksDB WAL and later flushed to disk.
+- ``bluestore_writing``: TBD
+- ``bluefs``:  file-system-like interface that provides just enough functionality to
+  allow RocksDB to store its “files” and share the same raw device(s) with
+  BlueStore.
+- ``buffer_anon``: stores arbitrary buffer data.
+- ``buffer_meta``: all the metadata associated with buffer anon buffers.
+- ``bluestore_cache_data``: mempool for writing and writing deferred.
+- ``bluestore_cache_onode``: meta information of object node stored directed in
+  bluestore cache.
+- ``bluestore_cache_meta``: key under PREFIX_OBJ where we are stored.
+- ``bluestore_cache_other``: right now account's for
+
+  - ``map_t``: we use it to track raw extents on disk for SharedBlob. We also use
+    it for the in-memory Blob with the blob namespace.
+  - ``coll_map``: collections_map
+  - ``csum_data``: checksum data
+- ``bluestore_cache_buffer``: accounts for buffer cache shards.
+- ``bluestore_extent``: a logical as well as physical extents, pointing to (some
+  portion of) a blob.
+- ``bluestore_blob``: in-memory blob metadata and associated cached buffers.
+- ``bluestore_shared_blob``: in memory shared blob state, stores reference of set of
+  collections it is part of (includes cached buffers).
+- ``bluefs_file_reader``: accounts for bluefs file reader buffer.
+- ``bluefs_file_writer``: accounts for bluefs file writer buffer.
+
+Check mempools usage
+~~~~~~~~~~~~~~~~~~~~
+
+Command to see BlueStore memory allocation in these mempools::
+
+     $ ceph daemon osd.NNN dump_mempools
+
+
+.. note:: see more:
+    https://github.com/ceph/ceph/blob/master/src/include/mempool.h