c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-22 10:37:35 +03:00
Code Issues Packages Projects Releases Wiki Activity

Consider dnode_t allocations in dbuf cache size accounting

Browse Source 
Entries in the dbuf cache contribute only the size of the dbuf data to
the cache size. Attached "user" data is not counted. This can lead to
the data currently "owned" by the cache consuming more memory accounting
appears to show. In some cases (eg a metadnode data block with all child
dnode_t slots allocated), the actual size can be as much as 3x as what
the cache believes it to be.

This is arguably correct behaviour, as the cache is only tracking the
size of the dbuf data, not even the overhead of the dbuf_t. On the other
hand, in the above case of dnodes, evicting cached metadnode dbufs is
the only current way to reclaim the dnode objects, and can lead to the
situation where the dbuf cache appears to be comfortably within its
target memory window and yet is holding enormous amounts of slab memory
that cannot be reclaimed.

This commit adds a facility for a dbuf user to artificially inflate the
apparent size of the dbuf for caching purposes. This at least allows for
cache tuning to be adjusted to match something closer to the real memory
overhead.

metadnode dbufs carry a >1KiB allocation per dnode in their user data.
This informs the dbuf cache machinery of that fact, allowing it to make
better decisions when evicting dbufs.

Sponsored-by: Klara, Inc.
Sponsored-by: Wasabi Technology, Inc.
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Rob Norris <rob.norris@klarasystems.com>
Closes #15511
This commit is contained in:
Rob N
2023-11-18 08:25:53 +11:00
committed by GitHub
parent 6c6fae6fae
commit 92dc4ad83d
5 changed files with 102 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files
cmd/dbufstat.in
+11 -9
View File
@@ -37,7 +37,7 @@ import re
bhdr = ["pool", "objset", "object", "level", "blkid", "offset", "dbsize"]
bxhdr = ["pool", "objset", "object", "level", "blkid", "offset", "dbsize",
"meta", "state", "dbholds", "dbc", "list", "atype", "flags",
"usize", "meta", "state", "dbholds", "dbc", "list", "atype", "flags",
"count", "asize", "access", "mru", "gmru", "mfu", "gmfu", "l2",
"l2_dattr", "l2_asize", "l2_comp", "aholds", "dtype", "btype",
"data_bs", "meta_bs", "bsize", "lvls", "dholds", "blocks", "dsize"]
@@ -47,17 +47,17 @@ dhdr = ["pool", "objset", "object", "dtype", "cached"]
dxhdr = ["pool", "objset", "object", "dtype", "btype", "data_bs", "meta_bs",
"bsize", "lvls", "dholds", "blocks", "dsize", "cached", "direct",
"indirect", "bonus", "spill"]
dincompat = ["level", "blkid", "offset", "dbsize", "meta", "state", "dbholds",
"dbc", "list", "atype", "flags", "count", "asize", "access",
"mru", "gmru", "mfu", "gmfu", "l2", "l2_dattr", "l2_asize",
"l2_comp", "aholds"]
dincompat = ["level", "blkid", "offset", "dbsize", "usize", "meta", "state",
"dbholds", "dbc", "list", "atype", "flags", "count", "asize",
"access", "mru", "gmru", "mfu", "gmfu", "l2", "l2_dattr",
"l2_asize", "l2_comp", "aholds"]
thdr = ["pool", "objset", "dtype", "cached"]
txhdr = ["pool", "objset", "dtype", "cached", "direct", "indirect",
"bonus", "spill"]
tincompat = ["object", "level", "blkid", "offset", "dbsize", "meta", "state",
"dbc", "dbholds", "list", "atype", "flags", "count", "asize",
"access", "mru", "gmru", "mfu", "gmfu", "l2", "l2_dattr",
tincompat = ["object", "level", "blkid", "offset", "dbsize", "usize", "meta",
"state", "dbc", "dbholds", "list", "atype", "flags", "count",
"asize", "access", "mru", "gmru", "mfu", "gmfu", "l2", "l2_dattr",
"l2_asize", "l2_comp", "aholds", "btype", "data_bs", "meta_bs",
"bsize", "lvls", "dholds", "blocks", "dsize"]
@@ -70,6 +70,7 @@ cols = {
"blkid": [8, -1, "block number of buffer"],
"offset": [12, 1024, "offset in object of buffer"],
"dbsize": [7, 1024, "size of buffer"],
"usize": [7, 1024, "size of attached user data"],
"meta": [4, -1, "is this buffer metadata?"],
"state": [5, -1, "state of buffer (read, cached, etc)"],
"dbholds": [7, 1000, "number of holds on buffer"],
@@ -399,6 +400,7 @@ def update_dict(d, k, line, labels):
key = line[labels[k]]
dbsize = int(line[labels['dbsize']])
usize = int(line[labels['usize']])
blkid = int(line[labels['blkid']])
level = int(line[labels['level']])
@@ -416,7 +418,7 @@ def update_dict(d, k, line, labels):
d[pool][objset][key]['indirect'] = 0
d[pool][objset][key]['spill'] = 0
d[pool][objset][key]['cached'] += dbsize
d[pool][objset][key]['cached'] += dbsize + usize
if blkid == -1:
d[pool][objset][key]['bonus'] += dbsize