[DRBD-user] Linstor - snapshot mouting
robert.altnoeder at linbit.com
Mon Jun 17 10:47:51 CEST 2019
On 6/17/19 8:53 AM, Gábor Hernádi wrote:
> > and why it creates lv snapshot with k flag?
> The k flag is used for the lvchange command, not for creation. If an
> administrator does not just deactivate a VG or that specific LV but
> also sets the "activationskip" flag for our LV [...]
For whatever reason, there are flags that "activate" or "deactivate"
volumes or even volume groups, so LVM can pretend they don't exist
although they are actually there. Some commands may activate them, but
then there is another flag that tells an activate command not to
activate some of the VGs/LVs. And then there are flags that tell a
command to ignore the flag that would otherwise tell the command to
ignore the volume when activating volumes.
What's even funnier is that some versions of LVM have those flags and
parameters, but others don't, apparently some of them set the flags
automatically, some distributions set or clear the flags upon reboot,
and different versions of LVM either fail to activate the VGs/LVs,
because the flags are present, or their commands don't work at all,
because they don't know those flags and command line parameters.
And the best part is that apart from calling those external commands to
control LVM, there does not seem to be any sane API that an application
like LINSTOR could call directly instead, and even the various error
codes that those commands return if something goes wrong are either
ambiguous or not even documented.
Why this completely insane collection of counterintuitive flags was
implemented in the first place is beyond my imagination (also, it is
worth noting that the original implementation of LVM in IBM AIX does not
have any of them, and it works just fine), but judging by the amount of
trouble they have caused us during development, I personally believe it
must have been a deliberate act of sabotage.
More information about the drbd-user