mirror of
https://github.com/ceph/ceph
synced 2025-01-12 14:10:27 +00:00
osd: drop default map cache size from 500 -> 200
OSDs don't need this from each other anymore since they send a pg_interval_map_t with pg notify. This cache is only used during startup when loading things up (and 200 is still > the max advance of 150), and when clients appear with super old maps (pretty unusual unless they are idle for long periods). It should mean a pretty significant cut in memory... Signed-off-by: Sage Weil <sage@redhat.com>
This commit is contained in:
parent
27d1e4d41e
commit
1ff51a299d
@ -591,8 +591,8 @@ OPTION(osd_tier_default_cache_min_read_recency_for_promote, OPT_INT, 1) // numbe
|
||||
OPTION(osd_tier_default_cache_min_write_recency_for_promote, OPT_INT, 1) // number of recent HitSets the object must appear in to be promoted (on write)
|
||||
|
||||
OPTION(osd_map_dedup, OPT_BOOL, true)
|
||||
OPTION(osd_map_max_advance, OPT_INT, 200) // make this < cache_size!
|
||||
OPTION(osd_map_cache_size, OPT_INT, 500)
|
||||
OPTION(osd_map_max_advance, OPT_INT, 150) // make this < cache_size!
|
||||
OPTION(osd_map_cache_size, OPT_INT, 200)
|
||||
OPTION(osd_map_message_max, OPT_INT, 100) // max maps per MOSDMap message
|
||||
OPTION(osd_map_share_max_epochs, OPT_INT, 100) // cap on # of inc maps we send to peers, clients
|
||||
OPTION(osd_inject_bad_map_crc_probability, OPT_FLOAT, 0)
|
||||
|
Loading…
Reference in New Issue
Block a user