as on tiny devices ZFS fallbacks to non-cached writes (grants are
consumed too quickly) while formatting time doesn't depend on
device size with ZFS (which was the original reasoning for the limits).
also increase OST size as sometimes local testing with ldiskfs fails
due to lack of space.
Test-Parameters: trivial testlist=sanity-scrub mdscount=2 mdtcount=4
Test-Parameters: fstype=zfs testlist=sanity-scrub mdscount=2 mdtcount=4
Change-Id: I8aad6c39d23a1d4c8db07b76e9de7fa2a664b1e5
Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/37217
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Olaf Faaland-LLNL <faaland1@llnl.gov>
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
# use small MDS + OST size to speed formatting time
# do not use too small MDSSIZE/OSTSIZE, which affect the default journal size
# 400M MDT device can guarantee uninitialized groups during the OI scrub
-MDSSIZE=400000
-OSTSIZE=200000
+[[ $MDSSIZE < 400000 || "$mds1_FSTYPE" == ldiskfs ]] && MDSSIZE=400000
+[[ $OSTSIZE < 400000 || "$ost1_FSTYPE" == ldiskfs ]] && OSTSIZE=400000
# no need too many OSTs, to reduce the format/start/stop overhead
[ $OSTCOUNT -gt 4 ] && OSTCOUNT=4