On 2.11.22 13:51, Alexander Atanasov wrote:
Hello,
On 2.11.22 4:32, Kui Liu wrote:
[snip]
[LIU]: QUEUE_FLAG_STANDBY bit MUST be cleared here, as it is the only
place where
the ploop device recovers from standby mode. In our use case, once a
ploop device enters standby
mode, the userspace will initiate recovery by replacing the top
delta file without destroying
the device, which is why the bit is clear in 'repalce_delta' in vz7.
static int ploop_replace_delta(struct ploop_device * plo, unsigned long
arg)
{
...
if (plo->maintenance_type != PLOOP_MNTN_OFF)
return -EBUSY;
...
}
If that is used to clear the flag then how is the case when the error
occurs while in maintenace handled? I think we should address this too.
We've had a talk on the subject with LIU so for the record:
- standby flag only stops requests from userspace
- requests issued by maintenance are not affected
- userspace in case of EBUSY must retry until the maintenance is finished.
--
Regards,
Alexander Atanasov
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel
