John,
Thanks again for the help. I think that we are getting somewhere.
I did this test on a couple of my personal linux machines running reiserfs,
as well as ext2, and lastly the one with the raid array and 80mbit scsi2
card. I don't meen to post something that seems a bit unrelated, but I just
found it interesting.
Here are my results.
linux (w/reiserfs, 5400rpm ata33 ide) 12mb/s
linux (w/ext2, 5400rpm ata33 ide) 3.57mb/s
bsd system in question:
bsd (10k scsi3) 29mb/s
bsd (telenet raid, scsi2 interface) 1.6mb/s
not a typo. I did this four times and had an average of 1.6mb/s. That is
awful.
Do you feel it a problem to have both the scsi3 internal drive and the raid
on the same controller? The reason for this was that the internal drive was
the old one, and the raid has been recently added. The internal isn't
normally even mounted.
Thank you again,
Jeff
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]On Behalf Of John R. Jackson
Sent: Friday, April 06, 2001 6:40 PM
To: Jeff Heckart
Cc: [EMAIL PROTECTED]
Subject: Re: large filesystem problems
>...that was actually all that consisted of the log. Does that tell
>you anything?
It all fits together with the Amanda server only waiting 45 minutes but
the estimates taking much longer than that.
>Does the disk having a problem seem like the most logical from your
>perspective? This is about a $6-8,000 telenet 6 disk raid 5 array with 10k
>scsi drives and a 80mbit adaptec scsi2 card. The system is overall slow as
>a dog. This is odd to me because it is a thunderbird 900 with 512mb
memory.
>I cannot figure this whole thing out.
Well, I'd say there is something fundamentally wrong with the system if
there are multiple slowdown symptoms.
Try some simple dd's of the raw device and see what kind of timing you
get. For instance:
time dd if=/dev/rsda0h of=/dev/null bs=64k count=8192
That will move half a GByte, then "do the math" to see what kind of
performance you're getting (and/or adjust the numbers to move more/less
data). For instance, on one of the disks on my system it took 72 seconds,
so that's ~7.3 MBytes/sec.
I'll bet yours will show a problem. But I don't know enough about your
system (or your system type in general) to advise where to go after that.
All I can suggest is that you start eliminating/swapping pieces until
it behaves or you find the culprit.
You might also do some serious SCSI chain checkout. For instance, it's
my understanding you cannot have a narrow device after a wide device
in the same chain. That kind of thing. And termination is always a
likely culprit.
Good luck.
>Jeff
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
