Yes, I seriously couldn't think of a better title for this. I am having a hard time explaining this, so all I can do is kinda ramble. Both systems are running under a hardened selinux profile, and I haven't seen this issue on any of my other systems which use lvm. The issue goes like this: I want to create a new lv. This is standard and uncomplicated, so: # lvcreate -n jfstest2 -L+50G server WARNING: lvmetad is running but disabled. Restart lvmetad before enabling it! /dev/server/jfstest2: not found: device not cleared Aborting. Failed to wipe start of new LV. I cringe because this is the second time I've seen this and it is now an actual issue. At some point during the debugging process, I decide to do real debugging. Increase verbosity / debug flags, strace it, etc. I then find that *this works*: strace -f lvcreate -n jfstest2 -L+50G server It isn't ptrace() itself that makes it happy, as it works if I feed it a lot of verbosity/debug flags. My nearest hunch is that the increased lag of the debug output is enough to let this function, but that's only a guess. Would love to get this squished.
...and I just remembered what it takes to sidestep and isolate this bug: network2 ~ # lvcreate -n domains -L+100G server -v -Zn DEGRADED MODE. Incomplete RAID LVs will be processed. Setting logging type to disk Finding volume group "server" Archiving volume group "server" metadata (seqno 64). Creating logical volume domains Creating volume group backup "/etc/lvm/backup/server" (seqno 65). Activating logical volume "domains". activation/volume_list configuration setting not defined: Checking only host tags for server/domains Creating server-domains Loading server-domains table (253:1) Resuming server-domains (253:1) WARNING: "server/domains" not zeroed Signature wiping on "server/domains" not requested Creating volume group backup "/etc/lvm/backup/server" (seqno 65). Logical volume "domains" created network2 ~ # lvs LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert domains server -wi-a----- 100.00g root server -wi-ao---- 150.00g network2 ~ # lvcreate -n domains2 -L+100G server -v -Zy DEGRADED MODE. Incomplete RAID LVs will be processed. Setting logging type to disk Finding volume group "server" Archiving volume group "server" metadata (seqno 65). Creating logical volume domains2 Creating volume group backup "/etc/lvm/backup/server" (seqno 66). Activating logical volume "domains2". activation/volume_list configuration setting not defined: Checking only host tags for server/domains2 Creating server-domains2 Loading server-domains2 table (253:2) Resuming server-domains2 (253:2) /dev/server/domains2: not found: device not cleared Aborting. Failed to wipe start of new LV. Removing server-domains2 (253:2) Creating volume group backup "/etc/lvm/backup/server" (seqno 67). network2 ~ # lvs LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert domains server -wi-a----- 100.00g root server -wi-ao---- 150.00g
(In reply to Eric Gisse from comment #0) > Yes, I seriously couldn't think of a better title for this. I can. Try again.
Also, it could be any of these: Bug #500732 Bug #525614 Bug #521960
(In reply to Jeroen Roovers from comment #3) > Also, it could be any of these: > > Bug #500732 > Bug #525614 > Bug #521960 Not without serious convincing, given that the problems exhibited in those bug reports are either relating to lvmetad or wholly unrelated to the issue. For some reason, that I cannot even begin to fathom, the lv creation process will fail unless I lag the hell out of it using debug/verbosity flags and/or strace. Or if I disable zeroing, as I have documented. Feel free to name it something more comprehensible, as well.