From 1a6346f43524937c9ea2c6b878c84ac293c114c9 Mon Sep 17 00:00:00 2001 From: Zac Dover Date: Sat, 17 Jun 2023 10:51:51 +1000 Subject: [PATCH] doc/rados: edit firefly tunables section Edit the firefly tunables section in /doc/rados/operations/crush-map.rst. Signed-off-by: Zac Dover (cherry picked from commit 66eb789d6ce3f38de70e731bcb7d92a307a95181) --- doc/rados/operations/crush-map.rst | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) diff --git a/doc/rados/operations/crush-map.rst b/doc/rados/operations/crush-map.rst index 293658196a024..5c37eea88aaea 100644 --- a/doc/rados/operations/crush-map.rst +++ b/doc/rados/operations/crush-map.rst @@ -805,11 +805,13 @@ Migration impact: firefly (CRUSH_TUNABLES3) ------------------------- -The ``firefly`` tunable profile fixes a problem with ``chooseleaf`` CRUSH rule -behavior. When this problem arises, PG mappings with too few OSDs can occur -when too many OSDs have been marked ``out``. +chooseleaf_vary_r +~~~~~~~~~~~~~~~~~ -The new tunable introduced by ``firefly`` is as follows: +This ``firefly`` tunable profile fixes a problem with ``chooseleaf`` CRUSH rule +behavior. This problem arose when too many OSDs were marked ``out``, which resulted in PG mappings with too few OSDs. + +This tunable, introduced in ``firefly``, is as follows: * ``chooseleaf_vary_r``: Whether a recursive chooseleaf attempt will start with a non-zero value of ``r``, as determined by the number of attempts the @@ -825,15 +827,14 @@ Migration impact: cause less data to move. straw_calc_version tunable --------------------------- +~~~~~~~~~~~~~~~~~~~~~~~~~~ There were problems with the internal weights calculated and stored in the -CRUSH map for ``straw`` algorithm buckets. More specifically, when there were -items with a CRUSH weight of ``0`` or with a mix of different and unique -weights, CRUSH would distribute data incorrectly (that is, not in proportion to -the weights). +CRUSH map for ``straw`` algorithm buckets. When there were items with a CRUSH +weight of ``0`` or with a mix of different and unique weights, CRUSH would +distribute data incorrectly (that is, not in proportion to the weights). -The new tunable introduced by ``firefly`` is as follows: +This tunable, introduced in ``Firefly``, is as follows: * ``straw_calc_version``: A value of ``0`` preserves the old, broken internal-weight calculation; a value of ``1`` fixes the problem. -- 2.39.5