Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 12 Dec 2000 13:38:18 -0800
From:      Kachun Lee <kachun@pathlink.com>
To:        Matt Dillon <dillon@earth.backplane.com>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: Extreme high load with 12/7 4-releng
Message-ID:  <200012122138.NAA69074@pathlink.net>
In-Reply-To: <200012122046.eBCKkUc52350@earth.backplane.com>
References:  <200012120230.SAA32402@pathlink.net> <200012121801.KAA42878@pathlink.net>

next in thread | previous in thread | raw e-mail | index | archive | help
At 12:46 PM 12/12/00 -0800, you wrote:
>
>:Yep, they are news (nntp) frontends, but no feed stuff. They run a custom
>:nntpd that access the spool by NFS. They have been running that way for
>:several years and the last rev to the nntpd was at least 6 months old. This
>:was the top from a server still running 4.1.1-releng (just before
>:4.2-release):
>:
>:--------
>:last pid: 73544;  load averages:  1.52,  2.07,  2.24    up 1+20:15:10
>:20:43:28
>:574 processes: 7 running, 567 sleeping
>:CPU states: 19.5% user,  0.4% nice, 22.9% system,  9.9% interrupt, 47.3%
idle
>:Mem: 286M Active, 92M Inact, 107M Wired, 15M Cache, 61M Buf, 996K Free
>:Swap: 600M Total, 600M Free
>:-------
>:
>:I saw there was a big different in Inact and no swap used.
>:
>:>:-----
>:>:last pid: 26893;  load averages: 36.05, 40.55, 47.01    up 0+04:29:07
>:>:18:14:54
>:>:513 processes: 9 running, 503 sleeping, 1 zombie
>:>:CPU states: 21.3% user,  0.7% nice, 30.0% system, 10.7% interrupt, 37.3%
>:idle
>:>:Mem: 196M Active, 204M Inact, 83M Wired, 17M Cache, 61M Buf, 1152K Free
>:>:Swap: 600M Total, 17M Used, 583M Free, 2% Inuse, 232K Out
>:
>:Here was a pxl from one of the servers... thanks for helping
>:
>:  UID   PID  PPID CPU PRI NI   VSZ  RSS WCHAN  STAT  TT       TIME COMMAND
>:    0     0     0   0 -18  0     0    0 sched  DLs   ??    0:00.74
(swapper)
>:...
>
>    Hmm.  It all looks normal, even the swap useage.  Except for the 
>    load average.  Is the machine less responsive then normal?  It could
>    just be the way the load average is being computed.
>
>						-Matt

Sorry to say, the system performance degraded with the high load. In
telnet, it felt like a system with load of 40. With the 'working' FreeBSD,
the servers would still performance better (load >10) with twice the
connections. I still suspect that it was related to the server start to
swap heavily even with 200+M of Inact memory? The strange part was the swap
Used stayed relatively low (>20M), but, with top, the swap In and Out
changed every refresh.

Regards

Best regards


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200012122138.NAA69074>