mbuf_cluster (FAIL SLEEP)

[Peter Wemm]

On 5/23/14, 6:18 AM, Peter B. Pokryshev wrote:
> Hi.
> Is it normal after 16 days of uptime:
>
> # vmstat -z
> ITEM                   SIZE  LIMIT     USED     FREE      REQ FAIL SLEEP
> ...
> 16 Bucket:              152,      0,      24,     101,     193,   0,   0
> 32 Bucket:              280,      0,      38,     102,     329,   2,   0
> 64 Bucket:              536,      0,      30,      33,     487, 142,   0
> 128 Bucket:            1048,      0,     997,      11, 6717030,17345735,   0
> ...
> mbuf_packet:            256, 12896820,    1449,    1646,9062649837,118865,   0
> mbuf:                   256, 12896820,    2193,    1762,17686258507,   0,   0
> mbuf_cluster:          2048, 2015128,    3095,    1793,26759484,241537,1100807
> mbuf_jumbo_page:       4096, 1007563,    2160,     864,2326876443,   0,   0
> mbuf_jumbo_9k:         9216, 298537,       0,       0,       0,   0,   0
> mbuf_jumbo_16k:       16384, 167927,       0,       0,       0,   0,   0
> mbuf_ext_refcnt:          4,      0,       0,       0,       0,   0,   0
>
> I mean 128 Bucket (FAIL) and mbuf_cluster (FAIL SLEEP)

Yes, this is normal and it doesn't mean what you might expect.  It's a 
generic failure counter, not an allocation failure counter.  eg: if an 
object that was just freed fails to fit in a per-cpu free items cache it 
counts as a "FAIL".

-Peter