Add support for rolling update of stateful Elastigroups #572
Conversation
The respawn command detects stateful Elastigroup and recycles stateful instances one at a time, instead of using the deploy API (that only works for the groups without stateful integrations).
|
No new tests were broken by this PR 😂 |
Add --batch-per-subnet for respawn of stateful Elastigroups
|
@a1exsh do you still want this change merged? If so please rebase the branch. The unit tests are fixed on master. |
|
@jmcs yes, we use the patched version and it saves us a lot of hassle. Thanks for fixing the tests! :) |
|
Issues
======
- Added 11
See the complete overview on Codacy |
| if batch_size is not None: | ||
| raise Exception("Batch size is not supported when respawning stateful ElastiGroups") | ||
|
|
||
| respawn_stateful_elastigroup(elastigroup_id, stack_name, region, batch_per_subnet, stateful_instances, spotinst_account) |
There was a problem hiding this comment.
Codacy found an issue: Line too long (128/100)
|
|
||
| if not stateful_elastigroup_ready(stateful_instances): | ||
| raise Exception( | ||
| "Stateful ElastiGroup {} is not ready: some instances are not in the ACTIVE state".format(elastigroup_id) |
There was a problem hiding this comment.
Codacy found an issue: Line too long (117/100)
| break | ||
|
|
||
|
|
||
| def recycle_stateful_elastigroup_instance(elastigroup_id: str, instance: dict, spotinst_account): |
There was a problem hiding this comment.
Codacy found an issue: Missing function docstring
|
|
||
| if batch_per_subnet: | ||
| instances_by_subnet = stateful_elastigroup_instances_by_subnet(region, stateful_instances) | ||
| for subnet, subnet_instances in sorted(instances_by_subnet.items(), key=lambda item: item[0]): |
There was a problem hiding this comment.
Codacy found an issue: Line too long (102/100)
| wait_for_stateful_elastigroup(elastigroup_id, spotinst_account) | ||
|
|
||
|
|
||
| def stateful_elastigroup_instances_by_subnet(region: str, stateful_instances: list): |
There was a problem hiding this comment.
Codacy found an issue: Missing function docstring
| @click.option( | ||
| "--batch-per-subnet", | ||
| is_flag=True, | ||
| help="Recycle ElastiGroup instances in batches per subnet. Valid only for stateful ElastiGroups.", |
There was a problem hiding this comment.
Codacy found an issue: Line too long (102/100)
| respawn_stateful_elastigroup(elastigroup_id, stack_name, region, batch_per_subnet, stateful_instances, spotinst_account) | ||
| else: | ||
| if batch_per_subnet: | ||
| raise Exception("Batch per subnet is not supported when respawning stateless ElastiGroups") |
There was a problem hiding this comment.
Codacy found an issue: Line too long (103/100)
|
|
||
|
|
||
| def respawn_stateful_elastigroup( | ||
| elastigroup_id: str, stack_name: str, region: str, batch_per_subnet: bool, stateful_instances: list, spotinst_account |
There was a problem hiding this comment.
Codacy found an issue: Line too long (121/100)
| return instances_by_subnet | ||
|
|
||
|
|
||
| def stateful_elastigroup_ready(stateful_instances: list): |
There was a problem hiding this comment.
Codacy found an issue: Missing function docstring
| @@ -1954,7 +1959,7 @@ def respawn_instances(stack_ref, inplace, force, batch_size_percentage, region): | |||
| ) | |||
| elif group["type"] == ELASTIGROUP_RESOURCE_TYPE: | |||
| respawn.respawn_elastigroup( | |||
| group["resource_id"], group["stack_name"], region, batch_size_percentage | |||
| group["resource_id"], group["stack_name"], region, batch_size_percentage, batch_per_subnet | |||
There was a problem hiding this comment.
Codacy found an issue: Line too long (106/100)
| while True: | ||
| time.sleep(WAIT_FOR_ELASTIGROUP_SEC) | ||
| act.progress() | ||
| stateful_instances = elastigroup_api.get_stateful_instances(elastigroup_id, spotinst_account) |
There was a problem hiding this comment.
Codacy found an issue: Line too long (105/100)
|
👍 |
1 similar comment
|
👍 |
The respawn command detects stateful Elastigroup and recycles stateful
instances one at a time, instead of using the deploy API (that only works for
the groups without stateful integrations).
Optional flag
--batch-per-subnetmay be used to recycle all instance of one AZ/subnet at a time.