RichP
NAXJA Forum User
- Location
- Effort, Pa
Need some help with a meta raid array on a solaris box. About at the end of my rope.
any meta device guru's here...
- Thread starter RichP
- Start date
Root Moose
NAXJA Forum User
- Location
- ON, Canada
ChiXJeff said:
Same here, sorry.
What's up with it? Maybe it'll job some brain cells over here.
RichP
NAXJA Forum User
- Location
- Effort, Pa
OK, got a Sun 280R, solaris 9, Sun Storedge A5200, 10 74gig drives in the front bay, 68gig on s0, 11 36 gig drives in the rear bay, 33gig on s6. Goal was two seperate raid5 meta devices, d0 and d1.
Created a metadevice d0, raid5 on 9 of the front disks, took about 8 hours to initialize, then newfs on the /dev/md/rdsk/d0, mounted it on /u02 as a mountpoint. Kicked off the second set of 10 drives in the back as metadevice d1 about 2 minutes after I kicked the first md off, interrupted it with a 'ctl-c' as it slowed the system down to a crawl as it was initializing at the same time as d0 so I figured I'd ssh in from the house and do it later. Then things went south and d1 showed as OK when I tried to metainit later on that nite but after 9 hours it had only completed .02%.
Tried to remove the metadevice with 'metaclear -f d1' I get 'waiting on /etc/lvm/lock' and thats where it will sit for hours. Deleted the lock file in /etc/lvm/ and it just recreated it a couple of minutes later.
All the md's are NOT mounted so they are not inuse, commented em out in vfstab incase rpc.mdb was doing an end run.
What I need to do is figure out how to manually delete the meta devices and start over with a clean slate w/o wiping the box. I also want to format the drives and have all the slices the same, using s0 on the front set was a bad idea, want to set them all up on s6. I'm also using /etc/lvm/md.cf not the tab file. Suggestions ?
Keep in mind that the md's are now part of the kernel so they are not to be toyed with if I understand them, once I found that file, md.conf in /kernel/drv, oh boy this is gonna be interesting.... never had had problem with these before, usually takes me a half hour to set this up, granted I don't normally configure a full house, I usually just concantate or stripe them. :laugh3:
Created a metadevice d0, raid5 on 9 of the front disks, took about 8 hours to initialize, then newfs on the /dev/md/rdsk/d0, mounted it on /u02 as a mountpoint. Kicked off the second set of 10 drives in the back as metadevice d1 about 2 minutes after I kicked the first md off, interrupted it with a 'ctl-c' as it slowed the system down to a crawl as it was initializing at the same time as d0 so I figured I'd ssh in from the house and do it later. Then things went south and d1 showed as OK when I tried to metainit later on that nite but after 9 hours it had only completed .02%.
Tried to remove the metadevice with 'metaclear -f d1' I get 'waiting on /etc/lvm/lock' and thats where it will sit for hours. Deleted the lock file in /etc/lvm/ and it just recreated it a couple of minutes later.
All the md's are NOT mounted so they are not inuse, commented em out in vfstab incase rpc.mdb was doing an end run.
What I need to do is figure out how to manually delete the meta devices and start over with a clean slate w/o wiping the box. I also want to format the drives and have all the slices the same, using s0 on the front set was a bad idea, want to set them all up on s6. I'm also using /etc/lvm/md.cf not the tab file. Suggestions ?
Keep in mind that the md's are now part of the kernel so they are not to be toyed with if I understand them, once I found that file, md.conf in /kernel/drv, oh boy this is gonna be interesting.... never had had problem with these before, usually takes me a half hour to set this up, granted I don't normally configure a full house, I usually just concantate or stripe them. :laugh3:
Root Moose
NAXJA Forum User
- Location
- ON, Canada
You're doing this on a "live" system? Tsk, tsk (C'mon, you knew that was coming). 
Gotta force those lusers into a realization of downtime for maintenance where the system is "unavailable" otherwise they take you for granted. Doesn't matter if you can do it in the background and while live. 
Hmm, off the cuff spidey-sense is telling me to go into format and/or ODS and wipe the partition table(s) and metadevices and recreate. Spidey-sense is also telling me there may be a command to update the image of what meta is supposed to know dynamically - might be the part of ODS I'm thinking about. Sorry, Solaris 6 was just coming out when I got out of Solaris enterprise stuff so I'm really rusty.
Can you schedule downtime for this box? That might be the best solution if the meta devices are in a weird state.
Does the A5200 have the latest prom stuff on it? I had a similar problem (slow building, weird states) on old A100s back "in the day" due to not most current firmware on the array hardware.
HTH
Gotta force those lusers into a realization of downtime for maintenance where the system is "unavailable" otherwise they take you for granted. Doesn't matter if you can do it in the background and while live. Hmm, off the cuff spidey-sense is telling me to go into format and/or ODS and wipe the partition table(s) and metadevices and recreate. Spidey-sense is also telling me there may be a command to update the image of what meta is supposed to know dynamically - might be the part of ODS I'm thinking about. Sorry, Solaris 6 was just coming out when I got out of Solaris enterprise stuff so I'm really rusty.
Can you schedule downtime for this box? That might be the best solution if the meta devices are in a weird state.
Does the A5200 have the latest prom stuff on it? I had a similar problem (slow building, weird states) on old A100s back "in the day" due to not most current firmware on the array hardware.
HTH
RichP
NAXJA Forum User
- Location
- Effort, Pa
I had this array up and running last week for the past 9 months on another system, also a 280R. This is whats driving me crazy, it was up and running last wed. The ODM does not come into play, the array is connected over a qlogic HBA [fiber] card. I think I fubar'd the metadb when I controld c'd the second raid array and it left it in some 'tween' state that I can't get out of with the meta commands. Thats why I need to find out the files that I need to 'zero' out and get back to stock before meta.
Root Moose
NAXJA Forum User
- Location
- ON, Canada
IIRC, there is a tool (cmd line and/or ODS) to "list" the contents of the metadb partitions/locations. I seem to remember in my mind's eye an ODS screen with a container that represents the metadb container.
If you were able to at least list that then you may be able to list, modify, write changes, etc.
Just throwing this out there - been a long time.
If you were able to at least list that then you may be able to list, modify, write changes, etc.
Just throwing this out there - been a long time.
Last edited:
RichP
NAXJA Forum User
- Location
- Effort, Pa
Bit the bullet and did a metaclear on d0, waiting for my prompt to come back in one window while I do a whoooooole bunch of ls's in another to see if the system just blew up....my toes are getting cramped from being crossed, sometimes working at home and wearing flipflops comes in handy...
RTicUL8
NAXJA Forum User
- Location
- Rocky Mountain High... Colorado!
I'd ask for pics, but you already bored me to sleep.... :speepin: :speepin: :speepin:
Root Moose
NAXJA Forum User
- Location
- ON, Canada
Yes! What colour flip-flips?RTicUL8 said:
