Whamcloud - gitweb
LU-11736 utils: don't set max_sectors_kb on MDT/MGT 16/34016/2
authorAndreas Dilger <adilger@whamcloud.com>
Fri, 11 Jan 2019 19:35:06 +0000 (12:35 -0700)
committerOleg Drokin <green@whamcloud.com>
Sat, 19 Jan 2019 06:29:46 +0000 (06:29 +0000)
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 <adilger@whamcloud.com>
Change-Id: I496603da13aae042f63cc37c0dea221a393ebbe5
Reviewed-by: Nathaniel Clark <nclark@whamcloud.com>
Reviewed-by: Wang Shilong <wshilong@ddn.com>
Reviewed-by: Gu Zheng <gzheng@ddn.com>
Reviewed-on: https://review.whamcloud.com/34016
Tested-by: Jenkins
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
lustre/utils/mount_utils_ldiskfs.c

index 02927ab..910e3b1 100644 (file)
@@ -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)