On Mon, Jun 12, 2017 at 02:37:48PM +0800, 于相洋 wrote: > Appreciate to receive you reply, Robin Hugh. > I have also try to adjust the vm configuration before ,but get no effect. > > Now I will accept your method to do "echo 2 "> drop_caches That's strictly a workaround. The OTHER way we can recover memory is to stop the OSD process for a given filesystem, umount the filesystem, (not just remount), mount it again and restart the OSD process. Can you share some of your OOM messages, and we can try and confirm if it's the same issue? Also possibly fixed in much newer kernels, but I think BlueStore is going to make the bug irrelevant as well by avoiding XFS systems with lots of inodes. In our case, each XFS filesystem has ~6M inodes (over ~52k directories), and this hugely impacts slab. >From "slabtop -s c", to sort by size: OBJS ACTIVE USE OBJ-SIZE SLABS OBJ/SLAB CACHE-SIZE NAME 4017576 1872371 46% 2.00K 251186 16 8037952K kmalloc-2048 19057016 17725140 93% 0.38K 455004 42 7280064K mnt_cache 4952967 2288054 46% 1.06K 167276 30 5352832K xfs_inode 5346579 2380752 44% 0.57K 191958 28 3071328K radix_tree_node 6701565 6694510 99% 0.10K 171835 39 687340K buffer_head 7478016 2597615 34% 0.06K 116844 64 467376K kmalloc-64 832624 256901 30% 0.50K 26021 32 416336K kmalloc-512 71220 68291 95% 3.50K 7916 9 253312K task_struct 238380 104089 43% 1.00K 7487 32 239584K kmalloc-1024 896658 612456 68% 0.19K 21349 42 170792K dentry 121930 116536 95% 0.61K 3616 52 115712K proc_inode_cache -- Robin Hugh Johnson Gentoo Linux: Dev, Infra Lead, Foundation Trustee & Treasurer E-Mail : robbat2@xxxxxxxxxx GnuPG FP : 11ACBA4F 4778E3F6 E4EDF38E B27B944E 34884E85 GnuPG FP : 7D0B3CEB E9B85B1F 825BCECF EE05E6F6 A48F6136 -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
