12-01-2017 03:03 AM
If you have a GMC that is highly available (consists of an active and passive node), and you decided that you want to promote it to a GM, which means that it needs to reboot.
Can you control the boot sequence so that if this GMC is providing a DNS service, it will continue to do so despite the reboot. What I am trying to achieve is service continuity by capaitilizing on the fact that this member is highly available.
Like can the active node reboot, which causes a failover to the passive node, then once the original active node is back online, the new active node can be rebooted, meaning that a failover will happen again to the original active node. All this happening without a DNS service outage. Can this be managed somehow.
Solved! Go to Solution.
12-04-2017 01:14 PM
While performing a GMC promotion, we can set a sequential notification to provide wait time for Grid members to join the new Grid Master. The sequential order for Grid members to join the new Grid Master begins with the old Grid Master and then the Grid members in FQDN order.The default delay time is 120 seconds. You can configure the delay time from a minimum of 30 seconds up to 600 seconds.
Regarding your current configuration:
1. Upon running "set promote_master" on the Grid Master Candidate, the request goes to the active node of the Grid Master Candidate
2. The active node restarts in the process of transitioning itself to the master state.
3. The passive node of the grid master candidate immediately assumes role of the grid master candidate (active node) since it has missed multiple consecutive heartbeats from the active node and since the active node of the grid master candidate is in the process of transitioning to the new grid master.
4. As soon as the previously active node of the grid master candidate comes back up, it comes back up as the new grid master.
5. The currrent active node of the Grid Master Candidate logs "restart: new master found" and restarts to come back up as the passive node of the new Grid Master.
6. All members, including the previous Grid Master should restart now (along with the above node) logging the same message "restart: new master found".
7. Though the previous Grid Master and other Grid Members are HA pairs, all nodes might restart simultaneously, because their active node would see the new master, log the message "restart: new master found" and even if their passive node transition to active, they would also see the new master and restart immediately. But you can stagger it as mentioned in the begining.
Hope this helps!
07-16-2018 03:31 PM
So if you stagger it, the old Grid Master "whether stand alone or HA pair" will restart to join the new Grid Master, then after 120 seconds "default" the first member in the FQDN order will restart then after another 120 seconds the next member in the FQDN order will restart. Is this correct.
04-02-2021 01:05 AM
is this one of Gird Topology limitations? because if i just have 2 boxes across 2 Data center, 1 for GM and 1 for GMC, when using promote to GM, all of systems about Grid manager / DNS / DHCP services will go down right? it's hard to accept.
have you had some advice for my scenario? except using HA pair because my network is running on layer 3.