John Goerzen
2013-Oct-25 13:46 UTC
Excessive RAM usage in btrfs_inode & btrfs_delayed_node
Hello,
This is an x86_64 box running 3.10. It is presently running restore from
the e2fs dump/restore, restoring a dump made on an ext4 system onto a btrfs
filesystem. The btrfs filesystem is the only btrfs filesystem mounted on
the machine, and it is mounted with only the noatime mount option. It
begins by creating millions of directories. During this process, I observed
strangeness in top:
KiB Mem: 8180424 total, 8033296 used, 147128 free, 356 buffers
KiB Swap: 4194300 total, 1941492 used, 2252808 free, 61740 cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ SWAP COMMAND
9551 root 20 0 6767m 5.3g 228 R 95.1 68.2 631:52.89 1.2g restore
9550 root 20 0 529m 29m 168 S 0.0 0.4 0:26.85 471m mbuffer
3916 root 20 0 146m 1424 668 S 0.0 0.0 0:00.28 340
NetworkManager
10534 root 20 0 126m 1612 560 S 0.0 0.0 0:01.07 16
udisks-daemon
4750 root 20 0 126m 772 432 S 0.0 0.0 0:00.05 824
console-kit-dae
3388 root 20 0 116m 564 140 S 0.0 0.0 0:00.20 692 rsyslogd
The system has 8GB RAM, and restore is using less than 7GB. The system is
otherwise idle and I had carefully plucked any other even slightly large
processes. Yet it is 2GB into swap and has very little RAM free. To
investigate that, I turned to slabtop, and found:
Active / Total Objects (% used) : 3058057 / 3117455 (98.1%)
Active / Total Slabs (% used) : 539669 / 539716 (100.0%)
Active / Total Caches (% used) : 121 / 212 (57.1%)
Active / Total Size (% used) : 1759683.27K / 1770895.45K (99.4%)
Minimum / Average / Maximum Object : 0.02K / 0.57K / 4096.00K
OBJS ACTIVE USE OBJ SIZE SLABS OBJ/SLAB CACHE SIZE NAME
1145183 1145183 100% 0.29K 88091 13 352364K btrfs_delayed_node
1145181 1145181 100% 1.02K 381727 3 1526908K btrfs_inode
478060 477416 99% 0.19K 23903 20 95612K dentry
135908 134151 98% 0.94K 33977 4 135908K ext4_inode_cache
45816 38564 84% 0.04K 498 92 1992K ext4_extent_status
28000 22874 81% 0.55K 4000 7 16000K radix_tree_node
21930 21633 98% 0.11K 645 34 2580K sysfs_dir_cache
20827 8033 38% 0.06K 353 59 1412K kmalloc-64
11872 11122 93% 0.03K 106 112 424K kmalloc-32
It seems there is something about btrfs_inode in particular that is
consuming a vast amount of RAM. Any ideas?
Thanks,
John
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html