On 10/02/15 17:56, OmegaPhil wrote: > On 10/02/15 14:08, OmegaPhil wrote: > >> I'll start fiddling with the boot scripts again now. > > > I have spent the last 4 hours repeatedly booting the machine, > rediscovering Debian's init.d scripts are called in parallel etc... > finally demonstrated that the problem is being caused by cgmanager. I'll > try to dig deeper next.
Finally... thanks to narrowing it down, I could recreate it on a VM and
reboot much faster. The invalid reference was key - when the aufs volume
is apparently successfully unmounted, cgmanager still 'keeps' it in its
mount namespace (reading cgmanager bugs gave me some ideas):
===========================================================
grep -ir 'bulk-storage' /proc/{1..9999}/mounts 2>/dev/null
===========================================================
This was a quick hack to detect any invalid mounts, and low and behold,
a single one from cgmanager:
===========================================================
/proc/1686/mounts:none /mnt/bulk-storage aufs
rw,relatime,si=adc2efa1f847f99f,create=mfsrr:4294967296,sum 0 0
===========================================================
I don't use anything like LXC or Docker yet, so I haven't launched any
containers/namespaces etc.
I will continue fighting this as a cgmanager bug, but any idea why aufs
doesn't know that the reference is still held?
Thanks
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------------ Dive into the World of Parallel Programming. The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/
