2.5.XX very sluggish

2.5.XX very sluggish

[Chris Morgan]

1.4Ghz Athlon via 82cxx chipset, software raid 1 scsi drives, currently 
running 2.4.21

With 2.5.73/74/75(the only ones I've tried thus far) the kernel boots fine 
until it tries to mount the reiserfs partition on the raid1 set.  Replaying 
the journal takes many times longer than with 2.4.  Once it gets past that 
point the whole machine appears to be quite sluggish.  Is this a known issue 
with reiserfs + software raid 1?  What information would be useful to aid in 
debugging?

I have smp disabled, preemption enabled, tried with generic x86 
disabled/enabled.

Thanks,
Chris


Please cc me on replies, I'm not on the list yet.

Re: 2.5.XX very sluggish

[Doug McNaught]

Chris Morgan <cmorgan@alum.wpi.edu> writes:

> 1.4Ghz Athlon via 82cxx chipset, software raid 1 scsi drives, currently 
> running 2.4.21
> 
> With 2.5.73/74/75(the only ones I've tried thus far) the kernel boots fine 
> until it tries to mount the reiserfs partition on the raid1 set.  Replaying 
> the journal takes many times longer than with 2.4.  Once it gets past that 
> point the whole machine appears to be quite sluggish.  Is this a known issue 
> with reiserfs + software raid 1?  What information would be useful to aid in 
> debugging?

What does 'hdparm' say about DMA settings on your drive under 2.5?

-Doug

Re: 2.5.XX very sluggish

[Chris Morgan]

Under 2.4.21:

For my boot drive, ide:

/dev/hdc2:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    =  8 (on)
 geometry     = 2100/255/63, sectors = 32740470, start = 996030

For the scsi drives(2 drives in a raid 1 setup):
/dev/sda:
 readonly     =  0 (off)
 geometry     = 17885/255/63, sectors = 287332384, start = 0

And some performance numbers, maybe these are helpful:

/dev/md1:
 Timing buffer-cache reads:   128 MB in  0.60 seconds =213.33 MB/sec
 Timing buffered disk reads:  64 MB in  2.22 seconds = 28.83 MB/sec

/dev/hdc:
 Timing buffer-cache reads:   128 MB in  0.56 seconds =228.57 MB/sec
 Timing buffered disk reads:  64 MB in  4.33 seconds = 14.78 MB/sec



Under 2.5.75:

/dev/hdc:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    = 256 (on)
 geometry     = 33483/16/63, sectors = 33750864, start = 0

These are with X running:

/dev/hdc:
 Timing buffer-cache reads:   128 MB in 30.58 seconds =  4.19 MB/sec
 Timing buffered disk reads:  64 MB in 20.98 seconds =  3.05 MB/sec
Hmm.. suspicious results: probably not enough free memory for a proper test.

/dev/md1:
 Timing buffer-cache reads:   128 MB in 108.76 seconds =  1.18 MB/sec
 Timing buffered disk reads:  64 MB in 44.86 seconds =  1.43 MB/sec
Hmm.. suspicious results: probably not enough free memory for a proper test.

With X shutdown:
/dev/hdc:
 Timing buffer-cache reads:   128 MB in 11.26 seconds = 11.36 MB/sec
 Timing buffered disk reads:  64 MB in  6.88 seconds =  9.30 MB/sec
Hmm.. suspicious results: probably not enough free memory for a proper test.

/dev/md1:
 Timing buffer-cache reads:   128 MB in 11.88 seconds = 10.77 MB/sec
 Timing buffered disk reads:  64 MB in  7.74 seconds =  8.27 MB/sec
Hmm.. suspicious results: probably not enough free memory for a proper test.

I'm running xf4.2.1 and the 'nv' driver.

Any ideas as to what else I can look at?

Thanks,
Chris


On Sunday 13 July 2003 12:45 pm, Doug McNaught wrote:
> Chris Morgan <cmorgan@alum.wpi.edu> writes:
> > 1.4Ghz Athlon via 82cxx chipset, software raid 1 scsi drives, currently
> > running 2.4.21
> >
> > With 2.5.73/74/75(the only ones I've tried thus far) the kernel boots
> > fine until it tries to mount the reiserfs partition on the raid1 set. 
> > Replaying the journal takes many times longer than with 2.4.  Once it
> > gets past that point the whole machine appears to be quite sluggish.  Is
> > this a known issue with reiserfs + software raid 1?  What information
> > would be useful to aid in debugging?
>
> What does 'hdparm' say about DMA settings on your drive under 2.5?
>
> -Doug

Re: 2.5.XX very sluggish

[Chris Morgan]

I see the same behavior with 2.6.0 test1.  I can't imagine I'm the only person 
seeing this issue, anything else I can do to help with debugging?

Chris


On Sunday 13 July 2003 11:21 pm, Chris Morgan wrote:
> Under 2.4.21:
>
> For my boot drive, ide:
>
> /dev/hdc2:
>  multcount    = 16 (on)
>  IO_support   =  1 (32-bit)
>  unmaskirq    =  1 (on)
>  using_dma    =  1 (on)
>  keepsettings =  0 (off)
>  readonly     =  0 (off)
>  readahead    =  8 (on)
>  geometry     = 2100/255/63, sectors = 32740470, start = 996030
>
> For the scsi drives(2 drives in a raid 1 setup):
> /dev/sda:
>  readonly     =  0 (off)
>  geometry     = 17885/255/63, sectors = 287332384, start = 0
>
> And some performance numbers, maybe these are helpful:
>
> /dev/md1:
>  Timing buffer-cache reads:   128 MB in  0.60 seconds =213.33 MB/sec
>  Timing buffered disk reads:  64 MB in  2.22 seconds = 28.83 MB/sec
>
> /dev/hdc:
>  Timing buffer-cache reads:   128 MB in  0.56 seconds =228.57 MB/sec
>  Timing buffered disk reads:  64 MB in  4.33 seconds = 14.78 MB/sec
>
>
>
> Under 2.5.75:
>
> /dev/hdc:
>  multcount    = 16 (on)
>  IO_support   =  1 (32-bit)
>  unmaskirq    =  1 (on)
>  using_dma    =  1 (on)
>  keepsettings =  0 (off)
>  readonly     =  0 (off)
>  readahead    = 256 (on)
>  geometry     = 33483/16/63, sectors = 33750864, start = 0
>
> These are with X running:
>
> /dev/hdc:
>  Timing buffer-cache reads:   128 MB in 30.58 seconds =  4.19 MB/sec
>  Timing buffered disk reads:  64 MB in 20.98 seconds =  3.05 MB/sec
> Hmm.. suspicious results: probably not enough free memory for a proper
> test.
>
> /dev/md1:
>  Timing buffer-cache reads:   128 MB in 108.76 seconds =  1.18 MB/sec
>  Timing buffered disk reads:  64 MB in 44.86 seconds =  1.43 MB/sec
> Hmm.. suspicious results: probably not enough free memory for a proper
> test.
>
> With X shutdown:
> /dev/hdc:
>  Timing buffer-cache reads:   128 MB in 11.26 seconds = 11.36 MB/sec
>  Timing buffered disk reads:  64 MB in  6.88 seconds =  9.30 MB/sec
> Hmm.. suspicious results: probably not enough free memory for a proper
> test.
>
> /dev/md1:
>  Timing buffer-cache reads:   128 MB in 11.88 seconds = 10.77 MB/sec
>  Timing buffered disk reads:  64 MB in  7.74 seconds =  8.27 MB/sec
> Hmm.. suspicious results: probably not enough free memory for a proper
> test.
>
> I'm running xf4.2.1 and the 'nv' driver.
>
> Any ideas as to what else I can look at?
>
> Thanks,
> Chris
>
> On Sunday 13 July 2003 12:45 pm, Doug McNaught wrote:
> > Chris Morgan <cmorgan@alum.wpi.edu> writes:
> > > 1.4Ghz Athlon via 82cxx chipset, software raid 1 scsi drives, currently
> > > running 2.4.21
> > >
> > > With 2.5.73/74/75(the only ones I've tried thus far) the kernel boots
> > > fine until it tries to mount the reiserfs partition on the raid1 set.
> > > Replaying the journal takes many times longer than with 2.4.  Once it
> > > gets past that point the whole machine appears to be quite sluggish. 
> > > Is this a known issue with reiserfs + software raid 1?  What
> > > information would be useful to aid in debugging?
> >
> > What does 'hdparm' say about DMA settings on your drive under 2.5?
> >
> > -Doug