Open fabbione opened 1 year ago
after a bunch of fencing events to test node recovery, I think the issue that causes servers to appear on different nodes is this:
Location Constraints: Resource: an-test-deploy1 Enabled on: Node: an-a01n01 (score:200) (id:location-an-test-deploy1-an-a01n01-200) Node: an-a01n02 (score:100) (id:location-an-test-deploy1-an-a01n02-100) Constraint: location-an-test-deploy1 Rule: score=-INFINITY (id:location-an-test-deploy1-rule) Expression: drbd-fenced_an-test-deploy1 eq 1 (id:location-an-test-deploy1-rule-expr) Resource: an-test-deploy2 Enabled on: Node: an-a01n01 (score:100) (id:location-an-test-deploy2-an-a01n01-100) Node: an-a01n02 (score:200) (id:location-an-test-deploy2-an-a01n02-200) Constraint: location-an-test-deploy2 Rule: score=-INFINITY (id:location-an-test-deploy2-rule) Expression: drbd-fenced_an-test-deploy2 eq 1 (id:location-an-test-deploy2-rule-expr) Resource: an-test-deploy3 Enabled on: Node: an-a01n01 (score:100) (id:location-an-test-deploy3-an-a01n01-100) Node: an-a01n02 (score:200) (id:location-an-test-deploy3-an-a01n02-200) Constraint: location-an-test-deploy3 Rule: score=-INFINITY (id:location-an-test-deploy3-rule) Expression: drbd-fenced_an-test-deploy3 eq 1 (id:location-an-test-deploy3-rule-expr) Resource: an-test-deploy4 Enabled on: Node: an-a01n02 (score:100) (id:location-an-test-deploy4-an-a01n02-100) Node: an-a01n01 (score:200) (id:location-an-test-deploy4-an-a01n01-200) Constraint: location-an-test-deploy4 Rule: score=-INFINITY (id:location-an-test-deploy4-rule) Expression: drbd-fenced_an-test-deploy4 eq 1 (id:location-an-test-deploy4-rule-expr) Resource: an-test-deploy5 Enabled on: Node: an-a01n01 (score:200) (id:location-an-test-deploy5-an-a01n01-200) Node: an-a01n02 (score:100) (id:location-an-test-deploy5-an-a01n02-100) Constraint: location-an-test-deploy5 Rule: score=-INFINITY (id:location-an-test-deploy5-rule) Expression: drbd-fenced_an-test-deploy5 eq 1 (id:location-an-test-deploy5-rule-expr)
Not sure which part of anvil creates those constraints tho.
I believe this has been resolved. If not, please let me know. If so, please close.
Still an issue.
* an-test-deploy1 (ocf::alteeve:server): Started an-a01n02
* an-test-deploy2 (ocf::alteeve:server): Started an-a01n01
* an-test-deploy3 (ocf::alteeve:server): Started an-a01n02
* an-test-deploy4 (ocf::alteeve:server): FAILED (blocked) [ an-a01n01 an-a01n02 ]
* an-test-deploy5 (ocf::alteeve:server): Started an-a01n02
Should be fixed by pr#408
* an-test-deploy1 (ocf::alteeve:server): Started an-a01n01
* an-test-deploy2 (ocf::alteeve:server): Started an-a01n02
* an-test-deploy3 (ocf::alteeve:server): Stopped (disabled)
* an-test-deploy4 (ocf::alteeve:server): Started an-a01n01
* an-test-deploy5 (ocf::alteeve:server): Started an-a01n01
Need to update to ensure that new servers are provisioned on the subnode with the most servers.
according to anvil design, all servers should (by default) be created and run on the same server. This is not the case.
Both with creation of a server and recovery, often happens that one of the test server is on a different anvil-node.
I have been able to trigger this issue both at create and recovery time.