Hide Forgot
Description of problem: I'm not sure if this only LVM fault. Here is my setup: Dell PE r815 with three controllers: H200, H700 and HBA 6Gb which connects to an external chassis and here is where a VG becomes lost, only this one VG. There are six VGs in the system in total, some of them comprise multiple devices, some just single device. VG which becomes lost spans multiple disks but only from that one external chassis. Every reboot renders: PVs .. cannot be found (or similar). It suffice to: $ vgscan --cache and: $ vgchange -ay lsi0 and VG comes back to normal... until next reboot. Version-Release number of selected component (if applicable): lvm2-2.02.130-5.el7_2.5.x86_64 How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Created attachment 1224389 [details] vanilla plain lvm.conf I believe
WARNING: Device for PV QFVVeZ-d1bh-zfGs-xkFE-ZZnz-T0c3-vqZ2DF not found or rejected by a filter. just updated the OS: lvm2-2.02.166-1.el7.x86_64 + 3.10.0-514.el7.x86_64 = problem persists
Hi Could you please attach 'pvs -vvvv' output when your VG is NOT visible ? Also 'lsblk' and whole boot log (ideally take lvmdump output) (eventually sosreport if RHEL user). My best guest is - you use 'lvmetad' and it's not informed about appearance of disk (your missing PV) in your system. If your PV sits on some 'special' disk type - it's quite well possible that udev rules are not handling well all events.
updates via yum and two reboots and system boots fine. But I am not saying lvm updates were the fix here, I've seen the system booted fine before the updates, one or twice, I don't know if it is random or there is a pattern to it. I'll try to reboot once a week and as soon as the problem occurs I'll get the info requested. One thing is certain, no special disks, it's Dell HBA to a Supermicro chassis with tradition PMR disks, Hitachi HDS723030ALA640 and Seagate ST6000NM0024-1HT17Z.
Closing this bug as requested trace is mandatory for getting any insight into this problem. We can't reproduce it without it. In case you can provide trace - please feel free to reopen.
ok