-
Suggestion
-
Resolution: Unresolved
-
None
-
19
-
Problem Definition
Currently, due to different limitations, it's hard to fully autoscaling feature (like in AWS or Azure) in Jira Datacenter environment.
Suggested Solution
Enable Jira DC to fully utilize all autoscaling options.
Workaround
Some manual steps are required, including but not limited to
- Manually copy the index before starting the node
- Manually cleaning the old nodes
- Checking the index status
- Checking the cluster.properties file
- is blocked by
-
JRASERVER-65538 Active nodes query for offline node messages and index operations
-
- Closed
-
-
JRASERVER-42916 Stale node ids should automatically be removed in Jira Data Center
- Closed
-
JRASERVER-66970 /status should indicate when indexes are broken on a node
- Closed
-
JRASERVER-67261 As a JIRA Datacenter Administrator I want to do an automated cold recovery from index a snapshot
- Closed
-
JRASERVER-68639 As an JIRA Administrator I want to see progress of index snapshot restore in UI
- Gathering Interest
- relates to
-
JRASERVER-69033 Add REST APIs methods to clean old node ids in JIRA Data Center
- Closed
- mentioned in
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
-
Page Loading...
Form Name |
---|
Upvote - the auto-scaling feature is not functional / practically does not exist, while the manual work needs to be done.
---------------
The point is, that Atlassian is using the hostname and not the local IP address as a registered connection in the cluster.
If that hostname can't be resolved (e.g. the logical FQDN), the node will not join the cluster and we will have a dead fish in the water (not a functional node).
Making the hostname approach for registering the node in the cluster is the weak spot in the auto-scaling process, as our example shows and I believe that we are not the only one.
The Ansible Playbook is gathering all EC2 metadata during the bootup and I believe you are using the value of the hostname from there as also you could use the IP address.
The backend of the problem:
The VPC has defined the DHCP Option Set which can't be changed because of business reasons.
The DHCP Option Set dictates the FQDN in the newly scaled-up node.
The (Linux) node is not set in the Ansible Playbook to join the domain from that DHCP Option set.
The installation and settings of the new node are made solo by the Atlassian Ansible Playbook, we are not in the position to do additional settings, all changes need to be done in the Ansible Playbook.
Does the Jira and Confluence Data Centre have the option in the cluster.properties file where can be set the use of the IP address over the hostname for the cluster register during the node boot-up process?
If does not, providing one will be the solution to this problem.