Global lock failed: held by other host

[liyuanchaoCh]

Hello, excuse me, how can I solve the "Global lock failed: held by other host " problem? How can I know which host holds the global lock?

[zkabelac]

You can try to play with 'lvmlockctl -i' command.

There is not very trivial way to tell 'which' other host holds the lock - important for a machine is to 'know' some other machine holds the lock.

To decode the host - you need to know 'mapping' - so in practice it might be quite similar to check which machine really owns the lock.

[liyuanchaoCh]

You can try to play with 'lvmlockctl -i' command.

There is not very trivial way to tell 'which' other host holds the lock - important for a machine is to 'know' some other machine holds the lock.

To decode the host - you need to know 'mapping' - so in practice it might be quite similar to check which machine really owns the lock.
So how should I deal with "Global lock failed: held by other host" to create vg with vgcreate?

[teigland]

It looks like you're using sanlock. You need to check each host to find the command that is holding the global lock. On each host run: "lvmlockctl -i" and "sanlock status". On the host with a command holding the global lock, those commands will show something like this:

$ lvmlockctl -i
...
LK GL sh ver 0 pid 18058 (vgs)

$ sanlock status
...
r lvm_ss:GLLK:/dev/mapper/ss-lvmlock:68157440:SH p 17794

This indicates that pid 17794 (a 'vgs' command) is holding the global lock.

The global lock should not be held for a long time, so perhaps the pid is hung while holding the lock. If you find a hung command, you might want to check what it's doing before killing it.

[liyuanchaoCh]

It looks like you're using sanlock. You need to check each host to find the command that is holding the global lock. On each host run: "lvmlockctl -i" and "sanlock status". On the host with a command holding the global lock, those commands will show something like this:

$ lvmlockctl -i
...
LK GL sh ver 0 pid 18058 (vgs)

$ sanlock status
...
r lvm_ss:GLLK:/dev/mapper/ss-lvmlock:68157440:SH p 17794

This indicates that pid 17794 (a 'vgs' command) is holding the global lock.

The global lock should not be held for a long time, so perhaps the pid is hung while holding the lock. If you find a hung command, you might want to check what it's doing before killing it.

Thanks. Does "GLLK" here mean holding global lock?

[teigland]

"GLLK" is the sanlock resource name that lvmlockd uses for the global lock. If sanlock status prints the line, it means that the lock is currently held.

(One minor correction to what I wrote in the previous comment: the pid of the vgs command displayed by lvmlockctl is 18058. The "p 17794" displayed by sanlock status refers to the pid of the lvmlockd daemon.)

[liyuanchaoCh]

"GLLK" is the sanlock resource name that lvmlockd uses for the global lock. If sanlock status prints the line, it means that the lock is currently held.

(One minor correction to what I wrote in the previous comment: the pid of the vgs command displayed by lvmlockctl is 18058. The "p 17794" displayed by sanlock status refers to the pid of the lvmlockd daemon.)

ok, thanks a lot