https://bugzilla.novell.com/show_bug.cgi?id=308557#c5
Stanislav Brabec changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
--- Comment #5 from Stanislav Brabec 2007-11-06 03:54:51 MST ---
I see two solutions:
1. Adding -d removable to the init script and adding a simple patch to ingore
later open() errors. Easy, prevents ugly warnings, but still no support for
hotplug.
2. Use suboptimal solution with hald-addon and some ugly hacks to prevent
multiple HUPs. (as optional feature - admins of systems with 1024 physical
discs wouldn't be happy even with one single rescanning of all discs after
adding new iSCSI volume)
3. Rewrite smartd to support dbus and device add/remove.
I am ready to do either 1. or 2. WONTFIX for 3., unless there will be an
explicit FATE/L3 request about it.
Bruce Allen wrote:
Hi Stanislav,
smartd has a '-d removable' flag that can be set. This prevents smartd
from exiting or issuing significant warnings if the disk is not found on
smartd startup. I think there have been patches sent to the support list
which extend this directive so that if open() fails then it is not treated
as serious. Could you inspect one of those patches and check it into CVS?
That would not require any external interaction such as sending HUP from
the hotplug monitoring layer.
Cheers,
Bruce
The problem is that DEVICESCAN only triggers one device scan, on smartd
start up.
There are two cases of interest. Here is how it *should* work in each
case:
CASE 1: hotplug device present when smartd started. In this case,
if open() fails then smartd should skip the device check but not exit or
otherwise complain. But smartd should continue to try open() on each
polling interval. If open() succeeds then the disk should be checked.
CASE 2: hotplug device NOT present when smartd started. In this case,
when the device is plugged in, one has to arrange for an external event
(HUP) to be sent to smartd so that it scans again.
Could you check the patch to be sure that it obeys correctly in CASE 1
above?
Cheers,
Bruce
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.