Drop the implementation of CHECK_SRVIDS control

[Amitay Isaacs]

Hi Volker,

On Thu, Aug 24, 2017 at 2:18 AM, Volker Lendecke <Volker.Lendecke at sernet.de>
wrote:

> Hi, Amitay!
>
> Can we get something similar back? I have a case where a smbd pid is
> re-used and ctdbd_process_exists marks a process as active where for
> the same pid stale records exist that need pruning. We need to re-add
> registering the unique id with ctdb and even possibly change the
> semantics of CHECK_SRVIDS: Make sure that the list of srvids given is
> attached to the same connection to make remote process checking
> reliable.
>

I think the current implementation of PROCESS_EXISTS control is wrong.
It returns the result of kill(pid, 0) without checking if the pid is a
client of
CTDB or not.  I don't see any reason why we want to check the status of
a random process on a node.

May be fixing process_exists implementation should be enough.  If you
still think we need CHECK_SRVID, I can add a new control.

Just another thought.  CTDB also has SERVER_ID related controls.
That might be well suited to identify each process registered with CTDB.
And treat SRVID as pure messaging construct.

Amitay.