Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!mit-eddie!genrad!decvax!ucbvax!uvacs.UUCP!uucp
From: uucp@uvacs.UUCP (UNIX-to-UNIX Copy)
Newsgroups: comp.os.vms
Subject: Submission for mod-computers-vax
Message-ID: <8704241016.AA08645@uvacs.cs.virginia.edu>
Date: Fri, 24-Apr-87 05:16:22 EDT
Article-I.D.: uvacs.8704241016.AA08645
Posted: Fri Apr 24 05:16:22 1987
Date-Received: Sun, 26-Apr-87 21:35:28 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 54

Path: uvacs!virginia!umd5!eneevax!mimsy!seismo!rutgers!ucla-cs!ames!ucbcad!ucbvax!ti-eg.CSNET"!"MCCORE::BOLTHOUSE
From: "MCCORE::BOLTHOUSE@ti-eg.CSNET"
Newsgroups: mod.computers.vax
Subject: State = RWMPB
Message-ID: <8704230948.AA06631@ucbvax.Berkeley.EDU>
Date: 22 Apr 87 14:47:00 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 43

Resource 
Wait 
Modified 
Page writer 
Busy:

This happens when the SWAPPER is trying to flush the modified list to
the [pagefile, swapfile, section file ... depending upon what kind of
page it is ... the swapfile is used if your pagefile(s) are too full].

If you have a program [say a Monte Carlo simulation, or a FORTRAN programmer
accessing 1000 * 1000 complex * 16 arrays in the wrong order :-( ] with
a working set small enough to cause paging, your modified list will quickly
reach MPW_HILIMIT/MPW_WAITLIMIT.  When this happens, the SWAPPER must recover
enough pages to allow use of the modified list again.  It does this by
writing pages.  Of course, this takes a while (I/O is expensive...).  If
a second process produces a modified page during this time, it can't be
added to the already-bloated modified list, so it goes into a resource
wait:  RWMPB.

It is especially nasty when you have a badly-behaved program on the
secondary processor, and the SWAPPER madly trying to write out pages on
the primary.  Mr. Rogers asks, "Can *you* say 'gee, my 8800/83xx appears 
to be hung'"?

We have had overly-zealous system managers perform @CRASH from the console
because the system got *so* badly glutted with modified pages it hung for
several seconds. Users then thought their processes were hung, so they did
the 'logical' thing: log in again.  Of course, with many users trying this
simultaneously, we exceeded BALSETCNT, which made the SWAPPER work harder,
which made it hang *even more*, and ... you get the picture.  Several processes
in RWMPB generally means your system has a problem.  Usually, this can be
traced to a badly-behaved application or poorly-balanced page caches (i.e.,
your modified list needs to be bigger!).

david l. bolthouse
texas instruments defense electronics information systems VAX system support

ma bell:  214.952-2059
arpa:     bolthouse%mccore@ti-eg.csnet
csnet:    bolthouse%mccore@ti-eg

disclaimer:  you know by now that my employer thinks i'm silly.