scale-out API will take a long time to response when too many nodes in a cluster
Bug #1677268 reported by
Ethan Lynn
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| senlin |
Fix Committed
|
High
|
Ethan Lynn | ||
Bug Description
I guess this is because we load the cluster object with all nodes, this will take time and sometimes slow down the response. And client will think it's a timeout because longer than 60s to get the response.
| Changed in senlin: | |
| status: | New → Triaged |
| importance: | Undecided → High |
| Changed in senlin: | |
| assignee: | nobody → Ethan Lynn (ethanlynn) |
Revision history for this message
| Duc Truong (dtruong) wrote | #1 |
| Changed in senlin: | |
| status: | Triaged → Fix Committed |
To post a comment you must log in.
This was probably due to other actions running and waiting for the cluster or node locks to be released. Previously if a lot of actions for the same cluster or node were started, they would all wait for the locks to become available. This slowed down the engine a lot.
Recently we changed the API implementation to reject requests when a cluster / node is already locked or has another conflicting action in progress, so this problem should not be happening anymore.