[sf-lug] the latest on the servepath colo story

[jim stockford]

Asheesh! outstanding!

    you're right--no redundancy. I like RAID 1 and
a backup system. On my little LAN I've got NFS
shares that I occasionally copy to (gotta automate
that, but sometimes it hurts when i think). they had
nothing at all.
    going bad sectors is the rule, I think, hence
inode and superblock redundancy. I can't tell
what happened in this case--i ended up putting
the corner of a heat sink on the 'y' key when I
saw what fsck was gonna do to my finger (got
a snack and came back to a big "yyyyyyyy..."
command, hit enter, command not found, all
totally sucked--750 empty files with inode
numbers in their names, under lost+found the
same, with occasional directories, a mess).

    how to deal with new bad sectors? <--------

    "total failure" == "disk don't work no more"
a new bad sector might threaten the occasional
file but not the entire drive--some drives work and
work, with increasing numbers of bad sectors,
just not MBR or related. right? (not right?)

    thanks lots for the description of your colo box.
    For me, one of the exciting things about Paul's
colo offer at ServePath is that we get to go to
the site to dick with the box and check out the
scene as a bonus. But that's me.

    I got a crick in my neck right after reading where
you wrote about Debian. Nathan is pitching
Debian, mainly by dismissing Red Hat, but still,
Nathan's a hot item and i gotta listen. We settled
on CentOS (built with Red Hat code), but with
future ambitions somehow to employ either
Debian or Ubuntu.



On Oct 17, 2006, at 4:04 PM, Asheesh Laroia wrote:

> On Tue, 17 Oct 2006, jim stockford wrote:
>
>>    i got a job recently on a box where the hard drive
>> went down--/boot, /, swap.
>>    It was a production box, and the owner was dying
>> to get his scripts and data back. it would have been
>> a lot easier if he'd had more partitions--then whichever
>> went bad, the others would be okay (except in the
>> case of total failure, which is pretty rare, yes?).
>
> I'll bite:
>
> Multiple partitions is no substitute for redundancy.  Did it fail due 
> to
> disk failure that RAID would have ameliorated?  Or user error that 
> backups
> would have saved?
>
> Also, in my (limited personal) experience, sometimes disks go bad 
> sectors
> at a time, and sometimes they just stop working; I wouldn't call total
> failure "pretty rare".
>
>> PS: i've heard the claim of making multiple swap parts
>> for performance. I'm betting that no one defends that.
>
> This is a great idea if you have two separate disks on which to tell 
> the
> kernel to swap.
>
> Sorry I can't be around for any more SF-LUG meetings, having moved 
> back to
> Baltimore to finish school.  But I did recently pick and order parts 
> for a
> rack-mounted server for $1800, and I might as well tell you guys about 
> it
> in case you want to cost-share among yourselves on a more exiting 
> (albeit
> not cheap!) machine:
>
> $220 - Core 2 Duo retail CPU with stock cooler
> $249 - Supermicro PDSMI+ motherboard
> $680 - 4 GiB of RAM (Crucial 533 Mhz DDR2 ECC unbuffered)
> $400 - 4x 400 GB Seagate SATA 7200.9 drives (on sale at outpost.com)
> $264 - Supermicro SE-813MT-300C 1U case
> $48  - AOC-IPMI20-E management card so we can telnet into the BIOS 
> remotely!
>
> The plan is to RAID-5 the disks for /var and /home and put / into a 
> RAID 1
> setup.  The Core 2 duo is really exciting because it's very energy
> efficient but also very powerful.  The specs are kind of high, but the
> idea is since none of us live near the colo center, we don't want to 
> have
> ot go back later and say, If only we had more {disk|RAM|CPU}.
>
> I'm putting this thing together as a personal project and sharing the 
> cost
> with a few other people so it's worthwhile, and it reminded me of what 
> you
> guys were talking about here.
>
> Of course, we're running Debian on it. (-:
>
> -- Asheesh.
>
> -- 
> Any reproduction or redistribution of the Software not in accordance 
> with the
> License Agreement is expressly prohibited by law, and may result in 
> severe
> civil and criminal penalties.
>
> _______________________________________________
> sf-lug mailing list
> sf-lug at linuxmafia.com
> http://linuxmafia.com/mailman/listinfo/sf-lug
>