Hi,
I was successful in building online mirroing system with two nodes
by drbd, and it seems to work nicely. But for me, there is one
question need to be clarified :
When one node is down, the other node acts as primary.
But in reality, the malfunction one possibly has serious
physical problem, e.g., cpu or hard disk is broken-down ..etc
and need to fix it for a long time. During this period, if the
single
node reboot the system once, then the partion can't be mounted.
(because the /dev/drbdxx is connected via socket.)
In this situation, I have to mount the partion manually and restart
the monitor service(httpd or sendmail for example), and meanwhile
there is no mirroring mechanism.
When the other node is ok, I need to resync it's partion, and bring
both nodes into initial environment. It is redtapery.
If there is any solution to handle this situation automatically ?
By the way, if I want to migrate the socket mechanism from kernel
level to user level(like ENBD does), and replicate the disk request
directly from filesystem level, is it possible ? Need to do a lot of
work ? :)
Best regards,
Jonathan Chen.
jonathan@example.com
I was successful in building online mirroing system with two nodes
by drbd, and it seems to work nicely. But for me, there is one
question need to be clarified :
When one node is down, the other node acts as primary.
But in reality, the malfunction one possibly has serious
physical problem, e.g., cpu or hard disk is broken-down ..etc
and need to fix it for a long time. During this period, if the
single
node reboot the system once, then the partion can't be mounted.
(because the /dev/drbdxx is connected via socket.)
In this situation, I have to mount the partion manually and restart
the monitor service(httpd or sendmail for example), and meanwhile
there is no mirroring mechanism.
When the other node is ok, I need to resync it's partion, and bring
both nodes into initial environment. It is redtapery.
If there is any solution to handle this situation automatically ?
By the way, if I want to migrate the socket mechanism from kernel
level to user level(like ENBD does), and replicate the disk request
directly from filesystem level, is it possible ? Need to do a lot of
work ? :)
Best regards,
Jonathan Chen.
jonathan@example.com