In short, the solution (i hope) is applying patch 105572-02.
Many thanks to all who responded. Responses are listed below.
On 20 March 1998 13:54 +0300, Serguei I. Revtov wrote:
> Hi All!
> I have continuous troubles with mirrored internal bootdisks on two
> UE4000s with Solaris 2.6 + Sun Enterprise Volume Manager 2.4
> installed. Sometimes VM cannot start rootdisk volumes so that system
> doesn't boot. :(
> There are next partitions on boot disk:
> * Partition Tag Flags Sector Count Sector Mount Directory
> 0 2 00 0 132240 132239 /
> 1 3 01 132240 2097600 2229839
> 2 5 00 0 4154160 4154159
> 3 7 00 2229840 267520 2497359 /var
> 4 15 01 4152640 1520 4154159
> 6 4 00 2498880 1229680 3728559 /usr
> 7 14 01 0 4154160 4154159
> Sure, i made only /,swap,/var and /usr partitions "by hands", other
> ones were made by VM.
> I have done the next steps:
> - install Solaris
> - install Recommended patches
> - install VM
> - apply patch 105208-03
> - reboot
> - vxinstall (with encapsulating boot disk)
> - reboot
> - make mirror on second internal disk (vxdiskadm)
> - reboot
> - create volume (+mirror) with remaining space on vxva.
> There are some volumes on SSAs which work fine...
> Everything work, but sometimes one of two servers can't boot because
> VM can't start /,swap,/var & /usr. Then i have to disable VM and
> reinstall it.
> That was tomorrow when i powered off/on both servers and they BOTH
> failed on booting :(
> Installed patches:
> 105208-03 105223-01 105357-01 105379-01 105407-01 105558-01
> 105216-01 105356-01 105375-03 105393-01 105518-01
> QUESTION is simple ;^)
> What should i do to make systems working for years?
> Do i need:
> - Go back to Solaris 2.5.1 or
> - Never mirror boot disks or
> - Get newer VM (2.5) or
> - Something else?
> Thanks in advance,
> i'll summarize.
> -- Serguei
>From: Leonid Massarsky (sysadm) <firstname.lastname@example.org>
Here's a fragment of README.105572-02 file of 105572-02 patch:
Topic: SunOS 5.6: /kernel/fs/ufs patch
4083720 Mirrored volumes resync on reboot, even when shut down
Hope it will help you. :-)
>From: Gert Marais <email@example.com>
I am surprised that 2.6 loads on Vm 2.4. When I upgraded to 2.6 my
machine would not load at all. I had to install VM 2.5 to be able to
load. I also mirror my internal system disks but I do offline mirroring
( Mirror only when something is O/S has changed and been tested ). I
personally will never encapsulate my root disks into volume manager.
What I also see is that your start and end sectors overlap which is the
first NO NO. Slices 4 and 6 does not look right and slice 7 is the
complete disk. I that is in volume managers control I do not even want
to think of what might happen. I would leave my O/S outside VM control.
If I would want to add the rest of the disk in Vm's control I would do
it from the command line where I can specify the slice of the disk to be
added. As far as I know vxdiskadm only allows the complete disk to be
added. Also read-up on encapsulating your root disk you might have
I'm having no problems with this setup.
>From: Kris Briscoe <firstname.lastname@example.org>
Personally I use Solstice Disk Suite to mirror my system disks, and use
VM to maintain my SSA's....Disk suite is much more simple to use for
system disks...and it works like a charm...I have never had a
problem....I would be glad to provide any details on it if you are
interested...If you have the server pack for your 4000, then SDS is part
of the package....
>From: bismark@alta.Jpl.Nasa.Gov (Bismark Espinoza)
The VM software is written so as to have an external disk array
which is managed by VM; and solaris disks inside the UE box.
If done this way, the disk array has to spin up before the
UE is booted.
>From: Misha Pavlov <email@example.com>
I had kinda similar problem with 2 internal 4G drives on ULTRA2 and 12 4G drives in multipack ( 14 4G drives total)
Volumes on Multipack were striped across drives and I had no problems.
C0t0d0 - 4G internal was encapsulated to rootdg and booting fine.
Any attempts to mirror c0t0d0 to c0t1d0 would fail and bring second internal drive to the state,
when even format would not be able to "inq" it.
Very weird ...
Came out to be that when u stuff Multipack full of drives you have to physically terminate it.
Moral - check cables and termination.
Thanks again to everyone who responded.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:34 CDT