From 8191b9b5a81a50edde7d5c949fb4c3aa7ed3f143 Mon Sep 17 00:00:00 2001 From: Andreas Dilger Date: Fri, 11 Jan 2019 12:35:06 -0700 Subject: [PATCH] LU-11736 utils: don't set max_sectors_kb on MDT/MGT The max_sectors_kb tunable should not be applied to MDT and MGT devices. This tuning is needed for efficiency of large IOs for spinning disks, but is not needed for SSDs or regular IO. It can cause problems with DM Multipath configurations for minimal benefits, so should be limited to OST devices. This only applies to ldiskfs backend filesystems, no such tuning is currently done for any ZFS devices. Lustre-change: https://review.whamcloud.com/33796 Lustre-commit: 2f8d7b4679de3fa467040aa61733f262714e39c9 Test-Parameters: trivial testlist=conf-sanity Signed-off-by: Andreas Dilger Change-Id: I496603da13aae042f63cc37c0dea221a393ebbe5 Reviewed-by: Nathaniel Clark Reviewed-by: Wang Shilong Reviewed-by: Gu Zheng Reviewed-on: https://review.whamcloud.com/34016 Tested-by: Jenkins Tested-by: Maloo Reviewed-by: Oleg Drokin --- lustre/utils/mount_utils_ldiskfs.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/lustre/utils/mount_utils_ldiskfs.c b/lustre/utils/mount_utils_ldiskfs.c index 02927ab..910e3b1 100644 --- a/lustre/utils/mount_utils_ldiskfs.c +++ b/lustre/utils/mount_utils_ldiskfs.c @@ -1116,6 +1116,13 @@ static int set_blockdev_tunables(char *source, struct mount_opts *mop) if (!source) return -EINVAL; + /* + * Don't apply block device tuning for MDT or MGT devices, + * since we don't need huge IO sizes to get good performance + */ + if (!IS_OST(&mop->mo_ldd)) + return 0; + ret_path = realpath(source, real_path); if (ret_path == NULL) { if (verbose) -- 1.8.3.1