osd: disable filestore_xfs_extsize by default

This option involves a tradeoff: When disabled, fragmentation is worse,
but large sequential writes are faster. When enabled, large sequential
writes are slower, but fragmentation is reduced.

Fixes: #14397
(cherry picked from commit aed85775bf53c273786ce4999320134822722af5)

diff --git a/src/common/config_opts.h b/src/common/config_opts.h
index f2b16c58bcb739d545bf04bf56d31c42b68541a9..25660a0bebb63ad34d1a1c72c5df2cf01c345c12 100644 (file)
--- a/src/common/config_opts.h
+++ b/src/common/config_opts.h
@@ -817,7 +817,10 @@ OPTION(filestore_fadvise, OPT_BOOL, true)
 // data corruption in xfs prior to kernel 3.5.  filestore will
 // implicity disable this if it cannot confirm the kernel is newer
 // than that.
-OPTION(filestore_xfs_extsize, OPT_BOOL, true)
+// NOTE: This option involves a tradeoff: When disabled, fragmentation is
+// worse, but large sequential writes are faster. When enabled, large
+// sequential writes are slower, but fragmentation is reduced.
+OPTION(filestore_xfs_extsize, OPT_BOOL, false)
 
 OPTION(filestore_journal_parallel, OPT_BOOL, false)
 OPTION(filestore_journal_writeahead, OPT_BOOL, false)