On Tue, 10 Apr 2007, Chung Hang Christopher Chan wrote:
Great, made your day then.
Not really, I use Solaris/OpenSolaris as my primary operating system.
This backup server was Open Solaris from the start
since I wanted to give Open Solaris a shot as there is
no iscsi target support included with the Linux kernel
as yet.
So, this would seem like a good attribute for Solaris.
Does Linux have good support for RAID bundled with
it?
RAID0, RAID1, RAID10 are okay. Never tried RAID5 since
I have heard quite a few horror stories about Linux
software raid arrays in raid5 mode.
What your experience on that though? What are you currently running your
backup server on, at your company?
What filesystem are you comparing zfs to?
Hmm. I don't believe I have actually made disparaging
comments about the Solaris kernel and stuff that comes
with it besides the userland packaging. And nobody has
done an in my face 'you do that with this on Open
Solaris' so I guess that means I have to wait.
Whoa there Chung Hang Christopher, no need to get defensive because I
asked what type of filesystems you compared it to. I was only trying to
find out if you were giving it a fair shake, before making your friend
laugh his pants off. Tell me what type of experience you have doing
backups. This sounds like something that one would use a server for, but
the majority of your comments are about the desktop lacking.
Ok, how did Solaris compare in the server environment? Is your friend
still laughing about Sun support? I have a friend who's stuck with Red Hat
support, and he's not laughing at all, he's crying...in his beer...
Ah, I don't know if you can find that in official
Redhat docs (must be somewhere in the RHEL3 docs...)
but you can go here for laughs:
http://www.nber.org/sys-admin/granite-digital-linux.html
The line: echo "scsi add-single-device 0 0 0 0"
/proc/scsi/scsi to tell the scsi system to rescan I
believe...
http://tldp.org/HOWTO/Software-RAID-HOWTO-4.html
The line: echo "scsi remove-single-device 0 0 2 0" >
/proc/scsi/scsi BEFORE you yank the disk...
Oh, and you have to get the numbers right too...
This is what I call "Google Support". Not only do we have that on Solaris,
you can go to docs.sun.com and get one stop shopping for much of the
system.
I was only talking about the hotswap procedures. :P
Well, did Solaris Express perform well in that regard?
regarding hardware raid...when will 3ware come to
Solaris? Are those guys not at all interested?
Have you already enquired with them? Which of their controllers are you
using?
Do you use ReiserFS? Ext3? What are you comparing
zfs to?
Both of those are a joke and so is XFS, the other
filesystem in the top three most popular Linux
filesystems, so I obviously did not make any ZFS vs
Linux 'x' filesystem.
Really? What filesystem do you run on Linux? Are you using Ext2? That's
getting pretty dated nowdays, it doesn't even support journaling, does it?
It's barely one step above FAT32...
You even have to wait for an fsck on Linux? Last I remember it would fsck
every 20 boots, odd way to maintain your system, but that's how I guess
it's done. How do you have your system configured to handle that?
How does your laughing friend manage that on his RHES? I bet he's not
laughing about that...<wink>
The two 750GB SATA disks for the mirror and they were
given whole without any partitioning.
So, you had 2 drives for the mirror, and another to boot off of?
Well, he does have another chipset, si3132 I believe.
Well, I wouldn't trust any other results on a different chipset,
especially something like a SATA driver, but it doesn't seem like you've
even gotten that far, as you're not sure if there's timeouts or not.
What software are you using for testing? IOW, what type of test are you
running this with?
My controller uses the si3124 chip. He is reporting cp
from one zfs to another zfs on pools sitting on sata
disks connected to his si3132 controller.
Different controller, would you expect your 3124 to be the same? Even revs
of NICs are not always supported, and you're using a SATA driver like
that?
I am
reporting rcp from a linux box (maildir mail host +
file server) to the solaris box will run into
'timeouts' at random times and i have apparently
worked around it by using rsync over rsh instead of
rcp
Did you file a bug on this? Although I would never reccomend any of the
'r' commands for general use, you're welcome to use and file bugs against
it when you find problems. Please do that if you want to make sure it gets
fixed. You can start at this link:
http://www.opensolaris.org/bug/report.jspa
In other words I wanted to rcp the initial stuff
over and then later use rsync to keep things in sync
but in the end I had to use rsync over rsh to the
initial copy since using rcp was too unpredictable.
This is with Solaris Express b59 so if there has been
an update to the si3124 driver/SATA framework, please
let me know how I can get them so I can test.
If you have a workaround, file the bug and use your workaround until it's
fixed. You can use scp on build 59 also, it's a part of the system. I
reccomend you use it rather than rcp.
--
Alan DuBoff - Solaris x86 IHV/OEM Group
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
