My original questions were:
Up front details: Solaris 2.4, Sparc1000, various SPARC clients,
one NCD X term.
1.Can a running process be reniced, and if so, how?
2.Can a running process be suspended and restarted?
And thanks for the replies from:
******************************************************************
From: Tim Hoffman <timh@momentum.com.au>
You could try running ufsdump with a much lower priority in the TS scheduler
using priocntl (nice doesn't work in Solaris 2.x)
Give the any process an exeedingly low priority and give
any process that must have good interactive performance
run in the INTERACTIVE scheduler.
This way any INT process will pre-empt any time share process.
I do this all the time so as to not bog a my 1+ when doing large compiles,
(the whole window system runs INTERACTIVE class, whilst I run back ground
compiles at a low priority in TIME SHARE.)
******************************************************************
From: Pell Emanuelsson <pell@lysator.liu.se>
> Can a running process be reniced, and if so, how?
Yes, with /usr/ucb/renice, although many people (me included) claim that
it has much less effect than a SunOS4 renice. For ufsdump you won't
notice any effect at all, since it's I/O bound.
> Can a running process be suspended and restarted? ( I tried sending
> the script process a kill -23 but top was still reporting the process
> running. Maybe I need to kill -23 all the instances of ufsdump started
Stopping the script won't help; you need to stop the ufsdump processes.
-23 and -25 would be the right way to accomplish this. Beware, though,
that it might cause other problems to stop the ufsdump, since it's first
mapping what to dump and then dumping. If you're dumping a live
filesystem, there should preferrably not be a large amount of time
passing from the mapping phase to the dumping phase.
******************************************************************
From: Kalpesh Dharia <kdharia@fh.us.bosch.com>
I'm not sure if this is what you're looking for. A running process can be
reniced. You can either use the renice or priocntl command for it. See
the man page for details.
I know it's not too precise, but I'm pulling it off my head. Hope it helps.
******************************************************************
From: freeman@sdt.com (Art Freeman)
It seems to me that running backups over a network would
increase the network utilization. Dataless and diskless
clients would probably be severely affected. In addition,
users running X apps over the network would also be affected
since X is bitmaped based unlike NeWS which is based on
display postscript.
******************************************************************
From: Herbert Wengatz <hwe@uebemc.siemens.de>
To me, it looks like you want to make a backup of your server-machine during
working hours. - Don't you never ever do that !!!!
Doing heavy I/O onto your SCSI-Subsystem will slow your whole Sun completely
down. - If you have any DL-clients or X11-Terminals connected to that machine,
their performance will hit immediately the ground hard. :-(
That's the reason every good system administrator makes his/her backups
automated during the night, when nobody is working.
******************************************************************
From: mark%mailhost@leav-156-100.army.mil (Mark B. Hamby)
A process can be reniced with the renice command or suspended with the
stop command.
The proctool program for Solaris 2.x is an excellant process management
and system resource tracking tool.
******************************************************************
From: Kevin.Sheehan@uniq.com.au (Kevin Sheehan {Consulting Poster Child})
The real solution here is probably not to be dumping with users on
the system. dump really wants a quite or unmounted FS anyway...
Kind of - see priocntl(1) for details.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:34 CDT