FC4 hang/interpret slabtop
ptfedora3 at majordomo.thedacare.org
ptfedora3 at majordomo.thedacare.org
Fri Jan 27 19:51:15 UTC 2006
I have a FC4 machine which hangs frequently.
It does a *lot* of logging from a remote syslog. It can hang at
unpredictable times, but will also usually become unresponsive around the
time when logrotate tries to rotate syslogs. My suspicion is there is
something wrong somewhere between the kernel and syslogd.
I have tried looking at slabtop to see if things are running out of room
in the kernel. I still think this is the case, but can't find any
authoritative source on increasing them to compensate.
Here is a view of my slabtop sorted by percentage used:
Things don't look right to me: there seem to be a lot of things at 100%
but my efforts to increase them by writing new values into /proc/slabinfo
are either wrong or ineffective. Any thoughts?
[root at syslogger ~]# slabtop
Active / Total Objects (% used) : 53671 / 74663 (71.9%)
Active / Total Slabs (% used) : 3335 / 3337 (99.9%)
Active / Total Caches (% used) : 85 / 123 (69.1%)
Active / Total Size (% used) : 9067.80K / 13123.39K (69.1%)
Minimum / Average / Maximum Object : 0.01K / 0.17K / 128.00K
OBJS ACTIVE USE OBJ SIZE SLABS OBJ/SLAB CACHE SIZE NAME
150 150 100% 0.12K 5 30 20K kmem_cache
3164 3164 100% 0.03K 28 113 112K size-32
195 195 100% 0.25K 13 15 52K size-256
160 160 100% 1.00K 40 4 160K size-1024
176 176 100% 2.00K 88 2 352K size-2048
161 161 100% 4.00K 161 1 644K size-4096
1 1 100% 8.00K 1 1 8K size-8192
1 1 100% 16.00K 1 1 16K size-16384
1 1 100% 32.00K 1 1 32K size-32768
1 1 100% 64.00K 1 1 64K size-65536
17 17 100% 4.00K 17 1 68K names_cache
26 26 100% 0.14K 1 26 4K sigqueue
77 77 100% 0.50K 11 7 44K sock_inode_cache
260 260 100% 3.00K 130 2 1040K biovec-(256)
14 14 100% 0.56K 2 7 8K UDP
570 570 100% 0.25K 38 15 152K ip_dst_cache
49 49 100% 0.50K 7 7 28K UNIX
32 32 100% 0.50K 4 8 16K sgpool-32
32 32 100% 1.00K 8 4 32K sgpool-64
32 32 100% 2.00K 16 2 64K sgpool-128
22 22 100% 0.34K 2 11 8K scsi_cmd_cache
8 8 100% 2.00K 4 2 16K rpc_buffers
265 264 99% 1.50K 53 5 424K biovec-128
434 431 99% 0.53K 62 7 248K shmem_inode_cache
59 58 98% 4.00K 59 1 236K pgd
275 272 98% 0.75K 55 5 220K biovec-64
2944 2884 97% 0.04K 32 92 128K sysfs_dir_cache
920 896 97% 0.04K 10 92 40K acpi_operand
280 272 97% 0.19K 14 20 56K biovec-16
1200 1140 95% 0.12K 40 30 160K size-128
84 79 94% 1.31K 28 3 112K task_struct
450 423 94% 0.44K 50 9 200K proc_inode_cache
81 76 93% 1.28K 27 3 108K sighand_cache
295 272 92% 0.06K 5 59 20K biovec-4
2508 2268 90% 0.09K 57 44 228K vm_area_struct
More information about the fedora-list
mailing list