Strange behavior of operating systems with lvm on Intel Modular Server

Question

linux

Good afternoon.

There is an intel Modular Server with 4 blades. On one of them, you need to deploy ubuntu with kvm and cars on lvm.
I install ubuntu from scratch (13.10) - it seems to be installed, lvm is created (default configuration, when /boot is outside lvm, and the rest is under lvm), but after reboot - busybox, lack of volume groups in /dev/mapper. The same with debian (Veasey). Centos (6.4) gets up, they didn’t try with LVM in /, then a separately created LVM partition was created on unallocated space, disks are also created through virt-manager, but when virtual machines are installed on them, they turn out to be read-only.

Advise the decision of this not clear situation. I really need ubuntu, I tried the solution from here: unix.stackexchange.com/questions/11125/lvm-devices-under-dev-mapper-missing, did not help. And in general, even the distribution behaves strangely, out of 5 times it sees hard drives only 1 time, the rest of the time it does not recognize them. The same with the Ubuntu live CD - for the first time I saw the disk and the volume groups created, and then nothing.

Answer 1

In debian, after installing the kernel, the installer asks which drivers to use for the initrd. All or some. Try to answer "all".
But in general it seems that he does not see the screws themselves, and not LVM on it.

Answer 2

Look when booting from the disk with ubunta - were there any warnings in dmesg regarding the controller? What's the driver? And make sure that this driver is either inside the kernel or inside the initramfs and loaded when it starts.
I had a case that I had to add pciid cards to new_id so that the driver would recognize it and start it, but, however, on the supermicro mother.

Answer 3

Actually, everything interesting in dmesge:

Fusion MPT base driver 3.04.20
Copyright (c) 1999-2008 LSI Corporation
Fusion MPT SAS Host driver 3.04.20
  alloc irq_desc for 42 on node 0
  alloc kstat_irqs on node 0
mptsas 0000:04:00.0: PCI INT A -> GSI 42 (level, low) -> IRQ 42
mptbase:ioc0:Initiating bringup
ioc0: LSISAS1064E B3: Capabilities={Initiator}
mptsas 0000:04:00.0: setting latency timer to 64
scsi0 : ioc0: LSISAS1064E B3, FwRev=01210000h, Ports=1, MaxQ=483, IRQ=42
mptsas: ioc0: add expander: num_phys 25, sas_addr(0x5001e67211fa52ff)
mptsas: ioc0: attaching ssp device: fw_channel 0, fw_id 0, phy 11, sas_addr 0x500015500002040a
scsi 0:0:0:0: Direct-Access Intel Multi-Flex 0310 PQ: 0 ANSI: 5
scsi 0:0:0:1: Direct-Access Intel Multi-Flex 0310 PQ: 0 ANSI: 5
sd 0:0:0:0: [sda] 251658240 512-byte logical blocks: (128 GB/120 GiB)
sd 0:0:0:1: [sdb] 524288000 512-byte logical blocks: (268 GB/250 GiB)
sd 0:0:0:0: [sda] Write Protect is off
sd 0:0:0:0: [sda] Mode Sense: 97 00 10 08
sd 0:0:0:1: [sdb] Write Protect is off
sd 0:0:0:1: [sdb] Mode Sense: 97 00 10 08
sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
sd 0:0:0:1: [sdb] Write cache: enabled, read cache: enabled, supports DPO and FUA
 sda:
 sdb:sda1 sda2
sd 0:0:0:0: [sda] Attached SCSI disk
 sdb1
sd 0:0:0:1: [sdb] Attached SCSI disk
dracut: Scanning devices sda2 for LVM logical volumes vg_server2/lv_swap vg_server2/lv_root
dracut: inactive '/dev/vg_server2/lv_root' [50.00 GiB] inherit
dracut: inactive '/dev/vg_server2/lv_home' [38.02 GiB] inherit
dracut: inactive '/dev/vg_server2/lv_swap' [31.48 GiB] inherit
EXT4-fs (dm-0): mounted filesystem with ordered data mode. Options:
dracut: Mounted root filesystem /dev/mapper/vg_server2-lv_root

From here we see that mptscsih does not swear in centosi, as in ubuntu. Just in case, I’ll explain that the driver is still mptsas, mptscsih is a library that is required in both cases and is called “Fusion MPT SCSI Host driver”. It's just that for some reason he does not swear in centos.
From the ubuntu log (above):

mptbase:ioc0:Initiating bringup
ioc0: LSISAS1064E B3: Capabilities={Initiator}
scsi0: ioc0: LSISAS1064E B3, FwRev=011e0000h, Ports=1, MaxQ=277, IRQ=30
mptsas: ioc0: add expander: num_phys 25, sas_addr(0x5001e67211fa52ff)
mptsas: ioc0: attaching ssp device: fw_channel 0, fw_id 0, phy 12, sas_addr 0x5000155000020409

FwRev is different, MaxQ is different (whatever that means). FwRev is, judging by the name, a version of the firmware, and in the log from the centos the number itself is larger - newer?
In this regard, the question is: did you receive these logs on the same machine?
By the way, looks like redhat had a similar problem four years ago: bugzilla.redhat.com/show_bug.cgi?id=483424#c12 . I have no idea if they succeeded or not.