Re: 3rd new wheezy install
On Wednesday, February 04, 2015 11:32:07 AM Chris Bannister wrote:
> On Wed, Feb 04, 2015 at 09:57:50AM -0500, Gene Heskett wrote:
> > convenience. This install is about the 6th I have done, and I am back to
> > the only way that works, which it let the SOB use all the disk as it
> > sees fit. And this is not a wheezy install as such, that install has
> > self destructed within 2 hours of the initial reboot while I was trying
> > to recover data that is in some cases 15 years old from the failing
> > drive.
>
> What do you mean here? How can you have 15 yr old data on a newly
> installed system?
>
The failing drive can still be read. Some of this data has now been on 10+
drives in the life of my use of linux, since 1998 TBE.
> > What is very discouraging in all this is that to a person, you _al_l
> > believe the installer can do no wrong, AND its onvious that this list is
> > not in any way connected to the people that build the install, so you
> > are not telling the install image people there is a problem with how it
> > handles these new 4k sector disks.
>
> You're going to have provide technical data. You must have worked on
> faults where the customer was saying on thing, but you knew you needed
> to check certain conditions yourself for diagnostic purposes.
>
That I have already posted, showing the output reports from other partitioners
say that thew partition boundary does NOT start on a sector boundary. The
partition boundary is normally considered to begin at any 512 byte sector on
the drive.
But these new high capacity drives ARE 4096 bytes per sector, so if the
partitioner is dumber than a rock, you only get the correct partition vs
sector 12.5% of the time.
And the partitoner that will not allow one to bypass its fscked up write on
partition tables IS DUMBER THAN A ROCK.
> > So the problem is not going to get fixed due to a lack of communication.
> ene@coyote:~$ sudo fdisk -l /dev/sda
> Correct, it's not. Hard data is what's required.
You are getting it! Here is an example from this operating drive right now:
$> sudo fdisk -l /dev/sda
[sudo] password for gene:
Disk /dev/sda: 1000.2 GB, 1000204886016 bytes
255 heads, 63 sectors/track, 121601 cylinders, total 1953525168 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 4096 bytes
Disk identifier: 0x000657ed
Device Boot Start End Blocks Id System
/dev/sda1 * 2048 1919977471 959987712 83 Linux
/dev/sda2 1919979518 1953523711 16772097 5 Extended
Partition 2 does not start on physical sector boundary.
/dev/sda5 1919979520 1953523711 16772096 82 Linux swap / Solaris
See that physical sector size? 4096 bytes. And the minimum I/O size? 4096
bytes.
Then note carefully what it says about partition 2, which is swap. Because of
that miss-alignment, swap will be even slower than it usually is. Linux CAN
work around it, but its at the cost of all that read-modify-write math it has
to use to compensate that in the swap address scheme, base 0 may be as far as
3584 bytes out of alignment with the beginning of that physical sector. Half
or more of the drives speed, plus at least 1 revolution of the platters after
the math and buffer modification has been done before it can write valid data
back.
How hard is that to understand, its a variation on a theme I've been fussing
about for a week.
Now, I can fix it, but the version of gparted I have, works in cylinder
boundaries, or if that is turned off, in mebibyte increments, neither of which
will automatically align on a 4096 byte per sector drive unless you start the
first partition 8 mebibytes into the start of the drive. I have tried 1, 2 &
4, which does not align. 8 does.
>
> > bug tracker isn't accessible to me as I don't even have incoming mail
> > setup such that I could confirm the subscription I'd need to file the
> > bug. Chicken v
>
> To be honest, a bug report in the manner you've reported here would not
> be at all helpful. You need to explain what you did, what you expected
> what the actual outcome was so that the person dealing with the bug can
> reproduce it using the same tools you used.
Thats the other point, I didn't do it, the installer forced me to do it its
way, there is no way to bypass it, even in the expert mode. That I tried 6 or
7 times, it simply will not allow you to proceed until it has written a
defective partition table, and then spends about an hour doing the mke2fs
thing on a miss-aligned drive. Something it should be able to do in 3 or 4
minutes on a terrabyte drive.
Now, if I have not explained it well enough for you, please show this message
to someone more familiar with the partitioner you use.
Just because you don't understand it, doesn't make it "not a problem". It is,
and the partitioner writers NEED TO BE TOLD there is a problem, and it needs
to be addressed since the 1st 4k/sector drives shipped almost 2 years ago.
Thirdly, the installer needs to be able to accept the partition table it finds
on the drive, format the SOB as it is and go on with the install. The
present, must partition the drive with your broken partitioner is not
acceptable when looking for best performance out of the drive.
I am not privy to enough info to do that, but that doesn't mean I'm your
average joe sixpack either. Neither do I feel like I have the time in what
remaining time I have left. Even with a tested IQ of 147, the wet ram is now
having problems of the short term memory variety, TBE when its 80 years old.
My bucket list is far from empty however. Re-writing your proprietary
partitioner is not on that list. I see all those names go flying by as its
loaded from the dvd, but none of that seems to be available in the repos.
Now, I don't have a sig generator setp yet, so
Cheers, Gene
Reply to: