[linux-lvm] Apparent race condition with srp and diskless machine

Mervini, Joseph A jamervi at sandia.gov
Tue May 2 20:26:20 UTC 2017


I am trying to use LVM2 with externally connected ip_srp storage hardware on a diskless booted node running RHEL 7 and when I attempt to create a logical volume or a snapshot under normal conditions I get the following error:

[root at ss53 ~]# lvcreate --snapshot --size 10G --name testsnap vg_local/lv_local-1 -v
  Using default stripesize 64.00 KiB.
    Setting chunksize to 4.00 KiB.
    Archiving volume group "vg_local" metadata (seqno 194).
    Creating logical volume testsnap
    Creating volume group backup "/etc/lvm/backup/vg_local" (seqno 195).
    activation/volume_list configuration setting not defined: Checking only host tags for vg_local/testsnap.
    Creating vg_local-testsnap
    Loading vg_local-testsnap table (253:57)
    Resuming vg_local-testsnap (253:57)
  /dev/vg_local/testsnap: not found: device not cleared
  Aborting. Failed to wipe snapshot exception store.
    Removing vg_local-testsnap (253:57)
    Creating volume group backup "/etc/lvm/backup/vg_local" (seqno 196).

(When I do the same operation with increased verbosity it shows that there is no device file being created in /dev/vg_local to support the new LV.)

The interesting thing is that when I run the command with -vvv at the console (via conman) it works! 

I don’t know whether there is a problem with lvm, udev or dmsetup that is causing the problem but it definitely seems like a timing issue. (If I do that same thing on a disk full system I don’t have any problems.)

Has anyone encounter this? Is there a work-around?

Thanks in advance.

Joe Mervini
Sandia National Laboratories
High Performance Computing
jamervi at sandia.gov

More information about the linux-lvm mailing list