Bill Stoddard wrote:
>
> This looks seriously hosed... There are clearly more than 89 requests being handled
>
> Current Time: Monday, 22-Oct-2001 10:26:59 PDT
> Restart Time: Monday, 22-Oct-2001 08:20:27 PDT
> Parent Server Generation: 2
> Server uptime: 2 hours 6 minutes 32 seconds
> Total accesses: 411687 - Total Traffic: 16.9 GB
> CPU Usage: u361.164 s597.625 cu626.703 cs184.898 - 23.3% CPU load
> 54.2 requests/sec - 2.3 MB/second - 43.1 kB/request
> 89 requests currently being processed, 11 idle workers
> W_WWWWW.KWWWWWWWWKWWWWWWCW_WWW_WC_WWW_WWWCWWWWWWW.WKCWWWWKWWW.KW
> KKKW_WWWWKWWWWWW_WKWKWW_WKWWW_KWWKW_WWCWWWWWWWWWW_WCWKWKWWWWWWWW
> WWKWWKWWWK_WWWWWWW.WWW_WWWKWWWWWWW_WWKWWWWKWKWWWWWGWWWK__WWW_KWK
> _WW_KWWCWWWWWKWW_WWWWWKWCWWWWWWWW_W_WWWKWKRKW_KCWWWKWW_WGWWWWW_W
> KWWWWWWKWWWCWWWWWWWWWWRWWWWKW_WWKW_WGWKWWWW_W_WKW_WWWWWWW_WWWW__
> WK_WWW_WWKWWWKWW_WWK_WWWWGWWKWWWWW._WWWWWWWWWK_WWR_WW_WWKKWWKCWW
> KW_WWW_K_WWWKWWKWWWWW_W_WKKK__K_WWCWWCWWKCWKWWWKWKWKWW__WWWWWWWW
> WWWWK.W.W....W...................W................._............
> ................................W...............................
For sure. It took me a long time to figure out that we were hitting
MaxClients earlier, because that number was so low.
To save people from counting, there are 64 state characters per line in
the status display above. I come up with 443 processes that aren't idle
('_') or dead ('.').
Perhaps we ought to provide a grand total of live workers, so an admin
can easily tell if there's a MaxClients problem, or whatever we end up
calling it. Then we could display subtotals by state along with the
"Scoreboard Key". We could save some room by not displaying unused
states. I've never seen some of those states.
There are some other bogusities in the status displays: child generation
number is 1 more often than it should be, negative numbers in the "SS"
column, "Req" column is always 0, and the "Conn" column looks dubious.
Greg
