Is maximum swap usage tunable?

[bob prohaska]

On Tue, Mar 06, 2018 at 11:28:17AM -0800, Rodney W. Grimes wrote:
> 
> Watch the output of the command I posted in the other
> thread,
> vmstat -z | egrep 'ITEM|g_bio|UMA'
> 
> There was a known g_bio leak.
> 

Here are a couple examples:

bob at www:~ % vmstat -z | egrep 'ITEM|g_bio|UMA'
ITEM                   SIZE  LIMIT     USED     FREE      REQ FAIL SLEEP
UMA Kegs:               408,      0,     199,       8,     199,   0,   0
UMA Zones:              520,      0,     200,       3,     200,   0,   0
UMA Slabs:               80,      0,     561,    1239,  119978,   0,   0
UMA Hash:               256,      0,       4,      11,      11,   0,   0
g_bio:                  376,      0,      16,      14,34532188,   0,   0
bob at www:~ % vmstat -z | egrep 'ITEM|g_bio|UMA'
ITEM                   SIZE  LIMIT     USED     FREE      REQ FAIL SLEEP
UMA Kegs:               408,      0,     199,       8,     199,   0,   0
UMA Zones:              520,      0,     200,       3,     200,   0,   0
UMA Slabs:               80,      0,    1172,     628,  120604,   0,   0
UMA Hash:               256,      0,       4,      11,      11,   0,   0
g_bio:                  376,      0,      32,     168,34710061,   0,   0


The very large integer in the REQ column looks odd, is it significant?

At roughly the same time, top looked like this:

last pid: 67582;  load averages:  3.48,  2.75,  2.50                        up 0+22:38:54  12:14:09
52 processes:  4 running, 48 sleeping
CPU: 88.0% user,  0.0% nice,  3.9% system,  2.5% interrupt,  5.6% idle
Mem: 625M Active, 48M Inact, 57M Laundry, 158M Wired, 84M Buf, 20M Free
Swap: 3072M Total, 827M Used, 2245M Free, 26% Inuse, 1016K In, 4308K Out

  PID USERNAME    THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
67533 root          1 102    0   543M   131M RUN     1   4:55  97.13% c++
67573 root          1 103    0   330M   302M RUN     2   1:45  96.64% c++
67582 root          1 102    0   140M   113M RUN     0   0:39  95.64% c++
67460 root          1  52    0   594M   115M swread  0   9:16  66.24% c++
  665 bob           1  20    0 13608K   892K CPU1    1   3:57   0.32% top
 7915 bob           1  20    0 12976K   512K nanslp  0   0:46   0.13% gstat
  652 bob           1  20    0 21008K   620K select  1   0:31   0.04% sshd
  571 root          1  20    0 15468K   736K select  1   0:08   0.02% sendmail
66462 root          1  20    0 16484K  2244K select  2   0:19   0.00% make
  667 root          1  20    0 10408K   208K select  0   0:09   0.00% make
  695 root          1  20    0 11096K   208K select  3   0:07   0.00% make
 7903 bob           1  20    0 21044K   320K select  2   0:06   0.00% sshd
20380 root          1  20    0 11008K   208K select  0   0:05   0.00% make
47745 root          1  20    0 14428K   216K select  0   0:04   0.00% make
47734 root          1  20    0 11000K   204K select  2   0:03   0.00% make
  578 root          1  20    0 11704K   216K nanslp  2   0:02   0.00% cron
56893 root          1  20    0 10408K   208K select  0   0:02   0.00% make

There were a couple of brief swread entries in the STATES column, but
the machine is generally making good progress; buildworld.log is up to
nearly 15 MB. I haven't noticed any pfault entries today. 

Of the five machines on my local net only two are being thoroughly 
exercised, a Pi2 and a Pi3. Both are running -current as of a couple
days ago. Both are having difficulty building world lately. The other 
machines (all Pi2) also build their own worlds and kernels but not often. 
They are running 11.

Thanks for writing!

bob prohaska