Koozali.org: home of the SME Server
Obsolete Releases => SME Server 7.x => Topic started by: ghopper on July 08, 2010, 07:31:37 PM
-
This has been a long road. I couldn't get 7.5.1 to install from either of two CDs I burned so I ended up doing a whole day of troubleshooting on the hardware with no positive result at the end. I gave up and installed 7.3 which is when I discovered the issue with the software RAID1 array. I tell you all this so you appreciate the frustration level. Oh, and as an aside, I'm no stranger to hardware, but most of my software experience is with Mac OSes and some PC, but I'm learning more about linux every day!
So, I had installed two identical 320 GB IDE drives in an older Dell Optiplex GX110. The drives had been used in a previous SME server install and then reformatted (on a PC with a cleanwipe program) to get rid of the old data. Install ver 7.3 and then I take a look at the disk redundancy status and this is what I see:
Personalities : [raid1]
md2 : active raid1 hda2[1] hdc2[0]
312464128 blocks[2/2] [UU]
md1 : active raid1 hdc1[0] hda1[1]
104320 blocks [2/2] [UU]
unused devices: <none>
all RAID devices are in a clean state
so, based on what reading I have done and what I THINK I understand, I believe that the 'c' drive is not properly partitioned/formatted(?) and so I need to somehow confirm that (and then presumably fix that.)
I do a bunch of reading about the fdisk command and cobble together what I think will tell me what I need to know:
fdisk -lu /dev/hda; fdisk -lu /dev/hdc
the result of which is:
Disk /dev/hda: 320.0 GB, 320072933376 bytes
255 heads, 63 sectors/track, 38913 cylinders, total 625142448 sectors
Units = sectors of 1 * 512 = 512 bytes
Device Boot Start End Blocks Id System
/dev/hda1 * 63 208844 104391 fd Linux raid autodetect
/dev/hda2 208845 625137344 312464250 fd Linux raid autodetect
Disk /dev/hdc: 320.0 GB, 320072933376 bytes
255 heads, 63 sectors/track, 38913 cylinders, total 625142448 sectors
Units = sectors of 1 * 512 = 512 bytes
Device Boot Start End Blocks Id System
/dev/hdc1 * 1 208769 104384+ fd Linux raid autodetect
Partition 1 does not end on cylinder boundary.
/dev/hdc2 208770 625137343 312464287 fd Linux raid autodetect
I'm guessing that partition warning is a sign that all is not good (but I'm still not sure if I am right.)
This is not a production machine yet so I presume that I can simply repartition the 'c' drive and then let it do it's thing in re-syncing the raid.????
I believe that I simply run fdisk on hdc, wiping out the old partitions, and then repartition (?)using numbers that duplicate hda (or can the software automatically sort that out?)
Beyond that I have two questions: 1) the start of hda1 (one of the apparently properly partitioned drives) is not at sector 1, but rather at 63. Why is this and is this correct?
2) the end of hda2 does not coincide with the number of sectors available on the drive (625142448) or even the total number of blocks. What gives?
And as I write this, I just noticed that somehow the software version has changed from 7.3 to 7.5.1 (update elves?) I did run the software installer last night before I went to bed so I am guessing that the 'updates' did more than update the packages, but actually updated the whole install to the latest and greatest. That was nice to find.
Am I on the right track with the repartitioning of the 'c' drive? and assuming that it will resync the whole thing once I'm done.
Thanks for any guidance on this problem!
ghopper
-
Install ver 7.3 and then I take a look at the disk redundancy status and this is what I see:
Personalities : [raid1]
md2 : active raid1 hda2[1] hdc2[0]
312464128 blocks[2/2] [UU]
md1 : active raid1 hdc1[0] hda1[1]
104320 blocks [2/2] [UU]
unused devices: <none>
all RAID devices are in a clean state
so, based on what reading I have done and what I THINK I understand, I believe that the 'c' drive is not properly partitioned/formatted(?) and so I need to somehow confirm that (and then presumably fix that.)
Why do you think that? Your system looks fully operational in RAID1 to me.
-
Thanks for responding!
I (apparently erroneously) believed that since the number of blocks on md1 and md2 were not identical, something wasn't right, but I wasn't 100% sure which led me to the fdisk examination.
When I saw the message regarding the partition issue, I was (again apparently erroneously based on your comment) much more sure I had discovered a problem.
So, no problem with the drives/partitions/RAID? And it appears I'm using the whole capacity of each drive?
Thanks again!
ghopper
-
I (apparently erroneously) believed that since the number of blocks on md1 and md2 were not identical,...
They are not expected to be. One is a small RAID1 array used for booting. The other is a large RAID1 array used for everything else.
-
Ok, I think I'm starting to get it. hda1 and hdc1 are the boot partition(s) and hold all the startup stuff (and the os?), (and it makes sense why it works this way - that they seem to be cross paired) and hda2 and hdc2 hold the 'rest' (storage, programs, etc..) Right?
So, why aren't hda1 and hdc1 identical? since they hold the exact same thing and are the exact same drive, it would stand to reason that they should be the same. Also, what do I make of the "Partition 1 does not end on cylinder boundary." warning? That's pretty much what I based my 'something is amiss' belief on.
Operating under the presumption that the numbers should be the same on an identical drive, I'm still wondering about the block 1 versus 63 starting point, and the differences in the total number of blocks (and ending point) of the second partition (as well as the issue of whether or not the whole drive is being used (which I suspect it is not.)
Lastly, I can't figure out what md1 (and md2) refer to. I'm guessing it is some sort of 'logical device' but I can't figure out what the 'md' stands for like I can with 'hd' and 'sd'. Any hints?
Thanks again,
ghopper
-
Lastly, I can't figure out what md1 (and md2) refer to. I'm guessing it is some sort of 'logical device' but I can't figure out what the 'md' stands for like I can with 'hd' and 'sd'. Any hints?
Multiple Device
http://linux.die.net/man/4/md
-
Thanks for the reference! I even answered another burning 'what does it stand for' question.
So, any chance you might comment on my other questions?
ghopper
-
http://lmgtfy.com/?q=linux+partition+%22sector+63%22
-
CharleyBrady I had tried to email you privately, but the site won't let me. I appreciate the Google search you offered, but at my knowledge level a search on Linux partitions "sector 63" doesn't seem to answer any of my questions about what is acceptable for a raid1 array. I'm assuming you aren't demonstrating 'scorn' for my lack of knowledge, but the search didn't address any of the top level questions I was asking (or thought I had asked.) It did direct me to the 'why of 63' and some other interesting stuff, but as best as I could tell, it didn't cover the meat of my questions (which I have broken out and amplified below.)
I also noticed that you only answer one question I ask out of each post and only with a single sentence. Unfortunately, I'm not getting a whole lot of learning out of those answers. I'm sure you want to help (as opposed to simply posting) and I'm guessing you might be working, where you might not have time to write a longer post that explains the what and why to my questions. So, thanks for your time; I need a bit more in terms of context or 'why' in the answers so that I 'learn to fish.'
I'm hoping someone can help me with the why of some questions. All questions reference the data as posted in my first post:
Question 1) background: two identical drives (but reused and formatted on a PC) configured in Raid1 with nothing more than the SME server install. md1 and md2 do not show an identical number of blocks (which I assume for lack of finding any other explanation of what the numbers represent) appears to mean that they do not have an identical amount of available storage. This seems like a problem since they are supposed to be identical drives and mirrors of each other. I understand that the difference may not be a problem, and in my case, it may not be an indicator of the problem.
I'm also wondering about the fact that the partition size is different from a to c as well (which may be an indicator of a problem.)
Q: Is the size difference of md1 and md2 a problem? If not, why not? specifically: a) what do the numbers represent, b) why aren't they identical (other than the fact that they start at different points), and c) depending on the answer to a and b, why isn't the difference a problem if they are supposed to be identical 'mirrors'?
Q: How come two identical drives have different sized partitions (hda1 vs hdc1 and hda2 vs hdc2)? It seems like hda1 and hdc1 should be the same size -even if a different starting block. Likewise with hda2 and hdc2) Should it be fixed? If so, what would the cleanest solution be? (right now I'm wondering about fdisking and reinstalling 7.3.)
Question 2) background: same two identical drives as above. hda1 and hdc1 do not start at the same sector. This seems like a problem since they are supposed to be identical drives and mirrors of each other. I keep coming back to the fact that hdc1, starting at sector 1 doesn't leave space for GRUB in the area before the partition (like David Bray explains, even to the point of showing and implying that everything should appear identical, and start at 63) which was why I was thinking that hda1 was set up correctly. However, in his tutorial and contrary to some of his statements, he appears to repartition, starting at 1 (not 63) but offers no explanation about the what/why of this change.
Q: So, repartition the drives and start hda1 and hdc1 at 1 or at 63? a) why one or the other? b) if it doesn't matter, why doesn't it matter (since it appears that the boot-loader is supposed to go in the section before 63?)
Question 3) Operating under the presumption that the numbers should be the same on an identical drive, I'm still wondering about the differences in the total number of blocks (and ending point) of the second partition as well as the issue of whether or not the whole drive is being used (which I suspect it is not.)
Q: Am I correct in my belief that the whole drive is not being used? should I do anything about that if it is not being completely used?
I have more but I'll stop there for now. Thanks in advance to anyone who is willing to be a teacher and help me understand the why of this stuff and what does and does not matter.
ghopper
-
Q: Is the size difference of md1 and md2 a problem? If not, why not?
I've already answered both questions, haven't I?
Q: Am I correct in my belief that the whole drive is not being used? should I do anything about that if it is not being completely used?
Yes, and no. Only a tiny proportion of the two drives is unused.
The difference in layout detail on the two drives is unlikely to be significant. I wouldn't worry about it unless you see a problem, or can discover something in your research which suggests there really is a problem.
-
this thread is where somebody helped me understand a bit more of it.
http://forums.contribs.org/index.php?topic=41408.0
Your DRIVES are HDA and HDC
Your raid arrays (volumes, partitions, whatever you understand) are MD1 and MD2
MD1 is the little boot array
MD2 is the array that has everything else
The [UU] indicates that the disks are mirrored and in sync.
If a disk fails you will see [_U] which indicates a degraded array.
-
ghopper
Re "Partition 1 does not end on cylinder boundary"
It's a non issue.
http://www.google.com.au/search?q=Partition+1+does+not+end+on+cylinder+boundary&ie=utf-8&oe=utf-8&aq=t&rls=org.mozilla:en-GB:official&client=firefox-a
and this answer for example
http://forums13.itrc.hp.com/service/forums/questionanswer.do?admit=109447627+1278736172927+28353475&threadId=1340438
Answered here too by Shad long ago
http://bugs.contribs.org/show_bug.cgi?id=2542#c1
"Linux is smart enough to manage the disk correctly."
-
Mary, thanks very much for the reply! Those links proved helpful, and working through all of this has been a great learning experience on both the server and linux in general.
The context sensitive advice is very helpful.
Last night I decided to try a little experiment and take the 'a' drive (hda) out. Wouldn't boot with the c drive. Being the good scientist that i am, I checked for false positives and put the a drive back in and took the c drive out. It booted. So, after putting the c drive back in, I went through the process of taking c out of the raid and repartitioning it, etc. (which was a great learning experience since I now get the logical relationships and am learning more about those tools (but what a brutal learning curve,))
Everything was good. Now, thanks to Mary's links, I even understand why I had a problem. The GX110 was an older machine (1999-2000) and several places I saw it mentioned that pre 2002 machines might have a problem (apparently due to 'that' particular machine's (really the bios') inability to do LBA at the bios level.) Since my spare box for this server is of the same vintage, this is good to know.
Thanks again for the guidance.
ghopper
-
Last night I decided to try a little experiment and take the 'a' drive (hda) out. Wouldn't boot with the c drive.
maybe a bios problem?
-
ghopper
...I decided to try a little experiment and take the 'a' drive (hda) out. Wouldn't boot with the c drive.
Try connecting the "c" drive to the "a" channel and then see if it boots.
-
What I actually did was take hda (on IDE#1) out and leave hdc in (on IDE#2). No boot. Swapped hda in to the same IDE channel (#2) that hdc was on (thinking that maybe the bios doesn't like booting off the #2 channel (though I know it will boot on #2 for the GX1, GX240 and so on.)) Booted just fine so it wasn't channel #2. Being the suspicious type, I put hdc on to #1 and no boot.
So, the problem followed hdc around.
After I read Mary's post and looked at the links. I discovered the info I alluded to in my earlier post: the bios in some computers manufactured before 2002 cannot handle partitions in larger drives that violate some of the older 'rules' regarding where a partition ends.
I haven't done a ton of reading about the why of the partition issue yet, but I have to travel on business this week so I may get to while I'm sitting around!
Next project is going to be getting the server to work with the outlook 2003 calendar and address book. Fortunately, there appears to be a huge amount of information out there on that topic.
Thanks again to all.
ghopper
-
What I actually did was take hda (on IDE#1) out and leave hdc in (on IDE#2). No boot. Swapped hda in to the same IDE channel (#2) that hdc was on (thinking that maybe the bios doesn't like booting off the #2 channel (though I know it will boot on #2 for the GX1, GX240 and so on.)) Booted just fine so it wasn't channel #2. Being the suspicious type, I put hdc on to #1 and no boot.
Please report this problem via the big tracker. Both disks should be bootable.