SOLVED: Strange Lag with 3.2

[John Eisenschmidt <jweisen_(_at_)_eisenschmidt_(_dot_)_org>]

I wanted to follow up and let everyone know that I think my mystery is
solved. I broke down and decided to replace my two crapola NICs with a
nice Zynx 4 port PCI NIC, and what a difference! Thank you to everyone
for your help on that.

As a slightly off-topic aside, I got a chance to run bonnie and
compare my previous 3Ware 6200 and mirrored IBMs to my new RAIDFrame
setup with mirrored 120GB WDs w/ 8MB cache:

3Ware Mirror:
-------------
              -------Sequential Output-------- ---Sequential Input--
--Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block---
--Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec
%CPU
          100  3891 18.0  3679  4.9  3217  4.7 16779 66.2 31220 26.6 175.8
1.4
real    2m3.246s

RAIDFrame Mirror:
-----------------
	      -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block---
--Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec
%CPU
          100 19838 38.1 19230 12.6  5857  3.9 12021 23.7 19536  8.2 181.9
1.7
real    1m8.776s

And thus, my upgrade from 2.9 to 3.2 is complete (just in time for 3.3
<G>). Thank you to everyone for all your hard work.


Rumor has it that John Eisenschmidt (jweisen_(_at_)_eisenschmidt_(_dot_)_org) said:
> Marco emailed me privately and suggested the following, my comments on
> how it went inline...
>
> ----- Forwarded message from Marco Peereboom <slash_(_at_)_peereboom_(_dot_)_us> -----
>
> > Reply-To: <slash_(_at_)_peereboom_(_dot_)_us>
> > From: "Marco Peereboom" <slash_(_at_)_peereboom_(_dot_)_us>
> > To: "'John Eisenschmidt'" <jweisen_(_at_)_eisenschmidt_(_dot_)_org>
> > Subject: RE: Strange Lag with 3.2
> > Date: Mon, 3 Feb 2003 14:39:39 -0600
> >
> > Try GENERIC with only the RAID driver. Don't use the NMBCLUSTERS,
> > APERTURE and the BUFCACHEPERCENT options.
> >
> > Aperture should be enabled in /etc/sysctl.conf line:
> > machdep.allowaperture=2         # See xf86(4)
> >
> > Try that and see what that does.
>
> Made the changes, built a new kernel, rebooted with the Intel fxp nic
> inside, same slowness within 5 minutes.
>
> > Then try to add your other options. Although I don't really see why you
> > need them.
>
> As I mentioned to Marco in a separate email, the bufcachepct and
> nmbclusters are to alleviate lag when playing mp3s over samba. Same
> problem with or without.
>
> > Good luck,
> > /marco
>
> Thanks to Marco and everyone who's responded so far, but still no
> luck.
>
> >
> > > -----Original Message-----
> > > From: John Eisenschmidt [mailto:jweisen_(_at_)_eisenschmidt_(_dot_)_org]
> > > Sent: Monday, February 03, 2003 13:28
> > > To: Marco Peereboom
> > > Subject: Re: Strange Lag with 3.2
> > >
> > >
> > > Rumor has it that Marco Peereboom (slash_(_at_)_peereboom_(_dot_)_us) said:
> > > > Shit, I did not even see that.
> > > >
> > > > Try GENERIC before you get flamed to death (maybe even
> > > deserved)! The
> > > > developers have no mercy with people complaining while
> > > using homemade
> > > > kernels.
> > >
> > > I would if I could -- RAIDFrame root mirror:
> > >
> > > jweisen_(_at_)_kaitain$ diff GENERIC kaitain
> > > 21c21
> > > < option                APERTURE        # in-kernel aperture
> > > driver for XFree86
> > > ---
> > > > #option               APERTURE        # in-kernel aperture
> > > driver for XFree86
> > > 26c26
> > > < #option               DUMMY_NOPS      # speed hack; recommended
> > > ---
> > > > option                DUMMY_NOPS      # speed hack; recommended
> > > 35a36,39
> > > > option                RAID_AUTOCONFIG
> > > > option                NMBCLUSTERS=8192
> > > > option                BUFCACHEPERCENT=30
> > > >
> > > 503c507
> > > < #pseudo-device        raid            4       # RAIDframe
> > > disk driver
> > > ---
> > > > pseudo-device raid            4       # RAIDframe disk driver
> > >
> > >
> > > > > -----Original Message-----
> > > > > From: owner-misc_(_at_)_openbsd_(_dot_)_org [mailto:owner-misc_(_at_)_openbsd_(_dot_)_org]
> > > > > On Behalf Of Peter Verhagen
> > > > > Sent: Monday, February 03, 2003 12:40
> > > > > To: jweisen_(_at_)_eisenschmidt_(_dot_)_org
> > > > > Cc: misc_(_at_)_openbsd_(_dot_)_org
> > > > > Subject: Re: Strange Lag with 3.2
> > > > >
> > > > >
> > > > > Is it possible for you switch back to the generic kernel and
> > > > > test pf again?
> > > > >
> > > > > Did you try a different PCI slot for the problematic NIC? For
> > > > > example, some mainboards with IDE RAID controllers tie
> > > into the same
> > > > > IRQ as one of PCI slots. As an example, my KR7A-RAID manual
> > > > > states:
> > > > >
> > > > > =-=-=-=-=-
> > > > > PCI Slot 5 shares IRQ signals with the HPT 372 controller
> > > > > (supports Ultra DMA133)....
> > > > > =-=-=-=-=-
> > > > >
> > > > > I've found that, even with the ATARAID disabled in the BIOS,
> > > > > any PCI cards in such a slot are flakey at best.
> > > > >
> > > > > Kind reguards,
> > > > > Peter Verhagen
> > > > >
> > > > > > As far as I can tell, yes. With the exception of the 2 new 120GB
> > > > > > drives (which I've had running for three weeks) all the
> > > > > hardware had
> > > > > > been in service at least a year without any problems. I really
> > > > > > hope
> > > > > > it's something I screwed up, and I really thought it was an
> > > > > interrupt
> > > > > > problem, but since I eliminated that this is the only thing
> > > > > I've come
> > > > > > up with, and I'm baffled. I don't know if it is the
> > > interraction
> > > > > > between raidframe and pf, but I can recreate the behavior after
> > > > > > disabling pf. The only combination I've come up with
> > > that doesn't
> > > > > > cause this problem is no Intel NIC on the inside.
> > > > > >
> > > > > > --
> > > > > > John W. Eisenschmidt (jweisen_(_at_)_eisenschmidt_(_dot_)_org)
> > > > > >   http://www.eisenschmidt.org/jweisen/pgp.html
> > > > > >
> > > > > > "If God were cheese, would we still worship on Sundays?"
> > >
> > > --
> > > John W. Eisenschmidt (jweisen_(_at_)_eisenschmidt_(_dot_)_org)
> > >   http://www.eisenschmidt.org/jweisen/pgp.html
> > >
> > > "The Religious Right is neither."
>
> ----- End forwarded message -----
>
> --
> John W. Eisenschmidt (jweisen_(_at_)_eisenschmidt_(_dot_)_org)
>   http://www.eisenschmidt.org/jweisen/pgp.html
>
> "Every program attempts to expand until it can read mail. Those programs
>  which cannot so expand are replaced by ones which can." -Zawinski's Law

--
John W. Eisenschmidt (jweisen_(_at_)_eisenschmidt_(_dot_)_org)
  http://www.eisenschmidt.org/jweisen/pgp.html

"If encryption is outlawed, then only outlaws will have encryption."

[demime 0.98d removed an attachment of type application/pgp-signature]