-
Improvement
-
Resolution: Unresolved
-
Minor
-
None
This plugin has really helped in the scenario of testing client-server, or cloud applications where we want to lock a deployment across multiple test jobs, however we end up duplicating configuration between lockable resources and node. A really useful improvement to lockable resources would be to choose the resource from online nodes in a pool represented by a node label.
For example, I currently configure:
- node label "BACKEND_POOL" representing a pool of Jenkins nodes where I can deploy my applcation back end/services
- lockable resources "SERVICE_POOL" with the identical list
Now I have a pipeline like pseudocode:
stage("Build") { build() } RESOURCE = get lockable resource from SERVICE_POOL stage ("Deploy") { node (RESOURCE) { deploy() } stage ("Test1") { node ("Test") { Test1(server=RESOURCE) } } stage ("Test2") { node ("Test") { Test2 (server=RESOURCE) } }
However, now I am maintaining the same list of resources in both the node label/pool and the lockable resource pool. This leads to issues like nodes being locked for deployment even though they are offline, someone adding or deleting a resource from one pool but not the other, or someone taking a node offline in one place but not the other.
Lockable resources should be able to lock a resource from online nodes in a node pool, and keep it locked across multiple stages that require using that resource but not as a node.
- is related to
-
JENKINS-44141 Define a resource pool
-
- Open
-
-
-
- Resolved
-
[JENKINS-53332] Allow picking a lockable resource from a node label
D Pasto
added a comment - JENKINS-44141 is talking about more complex data than the simple one to many relationship whereas this is integration with node labels/resources. There is nothing in that ticket that covers this, and they are not necessarily related but certainly could be implemented together.
D Pasto
added a comment - JENKINS-44141 is talking about more complex data than the simple one to many relationship whereas this is integration with node labels/resources. There is nothing in that ticket that covers this, and they are not necessarily related but certainly could be implemented together. my use-case would benefit from this as well: I have several nodes, each of which has a number of resources (of the same type) that I want to be able to be locked.
I also have nodes that do not have such resource.
When a job needs such resource, I actually need a node that is online, and has 1 free resource of the (common) label.
This is not possible right now. I solve this now by creating all the resources by including the node-name and a sequence number in the resource name. So eg for label 'PS-RESOURCE', I would create 4 lockable resources: PS-RS-nodeA-01,PS-RS-nodeA-02,PS-RS-nodeB-01,PS-RS-nodeB-02,
But this fails when a node goes off-line, as Jenkins has no coupling between the node and the resource. Although I would prefer a solution where the lockable resources are defined in the node configuration, I would also be happy if, for each lockable resource I would have the option to indicate a 'served by node' field which would make the resource unavailable if the node is offline.
I think this is a dupe/specialized case of JENKINS-44141.